Compare commits

...
188 Commits
Author SHA1 Message Date
Phil acd6bf819d Merge branch 'fix/v2.20.2-reply-threads' for v2.20.2 release (includes v2.20.1)
Build Container / build (push) Successful in 9m49s
2026-11-13 21:09:06 -07:00
PhilandClaude Opus 5.5 bf6e0fdc0b fix: backfeed whole reply threads and nest them (v2.20.2)
Backfeed only read first-level replies (Threads /replies, GoToSocial
direct replies, Bluesky depth 1) and skipped the owner's own replies, so
conversations under crossposts lost everything after the first reply.

- Threads /conversation, all GoToSocial descendants, Bluesky depth 10
- backfeed.thread_members keeps others' replies to our copy and any reply
  (ours included) to a kept reply; own replies are stored as notified
- Migration 026: webmentions.parent_remote_id; replies render nested as
  comments of comments

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-13 20:43:30 -07:00
PhilandClaude Opus 5.5 952ca300c0 fix: keep GIFs moving in reply contexts (v2.20.1)
Mastodon turns GIFs into silent looping MP4s (gifv); only the first-frame
preview was kept. The MP4 is now saved locally and played muted and
looping with the still as poster. Animated GIFs from other sites are saved
as animated WebP. Migration 025 adds reply_contexts.video and has
fediverse contexts with an image fetched again once.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-13 20:28:20 -07:00
Phil 03a1c8fcea Merge branch 'feat/v2.20.0-reply-crossposts' for v2.20.0 release
Build Container / build (push) Successful in 9m41s
2026-11-13 20:07:56 -07:00
PhilandClaude Opus 5.5 67c0411fd8 feat: crossposted replies look like replies (v2.20.0)
- Native replies on the same network: GoToSocial in_reply_to_id (resolved
  by our instance) with the author's @mention; Bluesky reply root/parent
- Replies to our own notes (permalink or any crosspost URL) thread under
  that note's crosspost on every network, Threads included
- Otherwise "↩ Replying to <author>" and the link; on Bluesky the name is
  the link. compose_text keeps the line whole when shortening

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-13 15:32:14 -07:00
Phil f7a936562c Merge branch 'feat/v2.19.0-reply-context' for v2.19.0 release
Build Container / build (push) Successful in 9m46s
2026-11-13 15:11:42 -07:00
PhilandClaude Opus 5.5 62e257e0e0 feat: reply contexts that show the post (v2.19.0)
Replies to Mastodon posts showed only the server's domain: the page has no
h-entry, so nothing was found and the host stood in for the author.

- Context sources in order: Bluesky AppView API, microformats, the
  origin's Mastodon API or this site's GoToSocial instance, OpenGraph
- Migration 024: author_handle, photo, photo_alt, source; rows without a
  source are fetched again once by the webmention worker
- Note page: a card with avatar, name, handle, text, image and date;
  home/tag/likes: "In reply to <author>" with a short quote
- Avatars and images saved locally (CSP allows only this site's images);
  avatar upkeep now covers reply contexts so their avatars aren't removed

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-13 15:10:46 -07:00
Phil 22e588377f Merge branch 'feat/v2.18.3-webmention-skip-hosts' for v2.18.3 release
Build Container / build (push) Successful in 9m38s
2026-11-12 13:42:10 -07:00
PhilandClaude Opus 5.5 b9e5321df2 fix: don't send webmentions to this site or WEBMENTION_SKIP_HOSTS (v2.18.3)
Links to the site itself, and to self-hosted apps like Gitea, timed out
from inside the container and sat in the admin Log. They are now skipped
when queueing; already queued or failed ones are dropped at worker start.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 13:40:27 -07:00
Phil 52ecf0f713 Merge branch 'feat/v2.18.2-tables-width' for v2.18.2 release
Build Container / build (push) Successful in 9m38s
2026-11-12 13:31:43 -07:00
PhilandClaude Opus 5.5 5fb521e89e style: header and footer share one 60rem frame on public and admin pages (v2.18.2)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 13:28:19 -07:00
PhilandClaude Opus 5.5 bba2178496 fix: tables fill the column as a card, matching common Markdown viewers (v2.18.2)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 13:24:37 -07:00
PhilandClaude Opus 5.5 41c373f46c fix: readable tables in notes and articles; public pages 48rem wide (v2.18.2)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 13:22:43 -07:00
Phil 7158713a2a Merge branch 'feat/v2.18.1-mcp-slug-docs' for v2.18.1 release
Build Container / build (push) Successful in 9m29s
2026-11-12 13:18:52 -07:00
PhilandClaude Opus 5.5 73a68157cc docs: MCP create_note describes article URLs and title slugs (v2.18.1)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 13:10:21 -07:00
Phil bda7adc449 Merge branch 'feat/v2.18.0-admin-cleanup' for v2.18.0 release
Build Container / build (push) Successful in 9m36s
2026-11-12 13:06:48 -07:00
PhilandClaude Opus 5.5 aa88db50b2 feat: admin Log and Settings pages, decluttered Webmentions page (v2.18.0)
Notifications move to /admin/settings. /admin/log lists everything failing
now (crossposts, response checks, outgoing webmentions, reply contexts,
WebSub, notifications, refused silo credentials) with the affected post,
the error and Retry/Check now. Back to Dashboard button removed; nav gains
Webmentions, Log (with error count) and Settings.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 13:05:26 -07:00
Phil c9b882aa46 Merge branch 'feat/v2.17.0-article-urls' for v2.17.1 release
Build Container / build (push) Successful in 9m8s
2026-11-12 12:53:59 -07:00
PhilandClaude Opus 5.5 6368365bd6 fix: build post URLs without a request context (v2.17.1)
Crossposting runs in a background worker; url_for there raised and every
crosspost failed. post_url now joins SITE_URL and post_path.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 12:53:54 -07:00
Phil 191c99b59b Merge branch 'feat/v2.17.0-article-urls' for v2.17.0 release
Build Container / build (push) Failing after 9m31s
2026-11-12 12:27:55 -07:00
PhilandClaude Opus 5.5 b7f8e1612c feat: articles at /article/<slug> with slugs from their titles (v2.17.0, ADR-077)
Articles (notes with a title) move from /note/<slug> to /article/<slug>;
the other prefix 301-redirects. New articles take their slug from the
title (mp-slug still wins; notes keep timestamp slugs). A Micropub update
that adds or removes a title answers 201 with the new Location.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 12:27:31 -07:00
Phil 66aecc2bcb Merge branch 'feat/v2.16.0-articles' for v2.16.0 release
Build Container / build (push) Successful in 9m35s
2026-11-11 17:25:39 -07:00
PhilandClaude Opus 5.5 ccb9f6dce2 fix: keep inline images in Micropub articles (v2.16.0)
- Content sent as both html and text is stored as the HTML; the text is
  usually a plain fallback, so images, links and headings were lost.
- Images embedded from other sites are copied into the media library
  (SSRF-checked fetcher, best effort, max 20) and the text points at the
  copies, on create and on content replace.
- HTML content is sanitized as is instead of going through Markdown, which
  added stray <br>s and empty paragraphs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-11 17:11:46 -07:00
PhilandClaude Opus 5.5 978c13358b feat: article titles and summaries stored apart from the text (v2.16.0, ADR-076)
Articles as the IndieWeb defines them: a name separate from the content.

- Migration 023 adds notes.title and notes.summary. A one-time Python data
  step (migrations.DATA_MIGRATIONS) moves each note's '# Heading' first line
  into title, rewrites the file and hash, empties FTS for a rebuild, and puts
  files back if the migration fails. Trashed notes included.
- Micropub: name -> title, summary stored when sent with content, summary
  updatable, q=source returns both. MCP tools take a summary.
- Feeds: titles only for articles (RSS/JSON none, Atom empty); Atom and
  JSON Feed summaries; the title is no longer repeated in the content.
- Pages: p-name/p-summary on the article page. Homepage and tag pages show
  articles as title, summary and "Read more" without images (hidden u-photo
  keeps the post type for parsers).
- Crossposts: articles go out as title, summary, link and hashtags, no
  images; the Bluesky card is the article. Notes unchanged.
- Editor: Title and Summary fields; an article may have no body.
- Fix: a line starting with a hashtag ('#indieweb ...') was rendered as a
  heading and crossposted without its '#' (HashtagLineExtension).
- Fix: excerpts showed raw Markdown; they are plain text now.

Back up the database and the data folder before deploying: the migration
rewrites note files.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-11 17:05:38 -07:00
Phil 9a0d82bab2 Merge branch 'feat/v2.15.0-mcp' for v2.15.0 release
Build Container / build (push) Successful in 9m40s
2026-11-11 16:39:39 -07:00
PhilandClaude Opus 5.5 84411a16d9 feat: MCP server for AI clients (v2.15.0, ADR-075)
Add an MCP server at SITE_URL/mcp so AI assistants (claude.ai connectors,
Claude Code) can search, read and write notes.

- starpunk/mcp_server: official MCP SDK 2.3, stateless Streamable HTTP with
  JSON responses, served by uvicorn as a sidecar from the same image.
  Ten tools; writes go through the Micropub handlers.
- IndieAuth: RFC 8707 resource indicators bound to codes and tokens
  (migration 022), client ID metadata document + public client metadata,
  draft in scopes_supported, default MCP scopes, consent screen note.
- BACKGROUND_WORKERS_ENABLED flag so the sidecar runs no workers.
- Docs: ADR-075, architecture/mcp.md and updates; CHANGELOG.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-11 16:36:58 -07:00
Phil 4ed2033cf9 Merge branch 'fix/v2.14.2-quotes-cleanup' for v2.14.2 release
Build Container / build (push) Successful in 9m37s
2026-11-10 15:54:55 -07:00
PhilandClaude Opus 5.5 d354dc9733 fix: quote counts in the counts line; remove dead editor scripts and files (v2.14.2)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 15:54:46 -07:00
Phil 88aea09e04 Merge branch 'fix/v2.14.1-tags-captions' for v2.14.1 release
Build Container / build (push) Successful in 9m6s
2026-11-10 15:37:43 -07:00
PhilandClaude Opus 5.5 7e2cdef150 fix: clear all tags from the editor; escape captions in feed HTML (v2.14.1)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 15:37:35 -07:00
Phil 058b1959f9 Merge branch 'docs/v2.14.0-doc-review': documentation review for v2.14.0 2026-11-10 15:35:00 -07:00
PhilandClaude Opus 5.5 ceef048003 docs: documentation review for v2.14.0
Architecture, README, standards, CLAUDE.md, settings reference, indexes,
backlog, release plans and CHANGELOG checked against the code at v2.14.0:
backfeed, notifications, counts line and web actions documented; test
policy (changed area locally, full suite in CI); stale versions, statuses
and links fixed; legacy and superseded docs moved.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 14:51:57 -07:00
Phil caa66f7bb1 Merge branch 'feat/v2.14.0-notifications' for v2.14.0 release
Build Container / build (push) Successful in 9m15s
2026-11-10 14:30:10 -07:00
PhilandClaude Opus 5.5 0465ff924b feat: notifications through Apprise (v2.14.0)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 14:30:01 -07:00
Phil 815aff5443 Merge branch 'docs/v2.14.0-notifications-adr': ADR-074 notifications 2026-11-10 14:25:58 -07:00
PhilandClaude Opus 5.5 4306633a42 docs: ADR-074 accepted
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 14:25:57 -07:00
PhilandClaude Opus 5.5 c5c2d9832e docs: ADR-074: one configurable priority for all messages
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 14:24:43 -07:00
PhilandClaude Opus 5.5 483c24b565 docs: ADR-074 notifications through Apprise (proposed)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 14:17:30 -07:00
Phil d312fa0074 Merge branch 'fix/v2.13.4-list-counts' for v2.13.4 release
Build Container / build (push) Successful in 8m56s
2026-11-10 14:05:05 -07:00
PhilandClaude Opus 5.5 647f6c723b feat: counts on lists of notes, reply cards link to the reply, fix class clash (v2.13.4)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 14:05:03 -07:00
Phil 81c730813c Merge branch 'fix/v2.13.3-count-opted-out' for v2.13.3 release
Build Container / build (push) Successful in 8m35s
2026-11-10 13:19:18 -07:00
PhilandClaude Opus 5.5 113f28aff1 fix: count opted-out likes and reposts without naming them (v2.13.3)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 13:16:19 -07:00
Phil 76b81d6ad9 Merge branch 'fix/v2.13.2-reaction-counts' for v2.13.2 release
Build Container / build (push) Successful in 8m34s
2026-11-10 13:11:30 -07:00
PhilandClaude Opus 5.5 44f77a3adf feat: likes and reposts as a counts line with hover lists, web actions (v2.13.2)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 13:11:29 -07:00
Phil b247de011d Merge branch 'fix/v2.13.1-avatar-fetch' for v2.13.1 release
Build Container / build (push) Successful in 8m33s
2026-11-10 12:56:11 -07:00
PhilandClaude Opus 5.5 65fc246d55 fix: save avatars from own GoToSocial and Bluesky PDSes (v2.13.1)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 12:56:04 -07:00
Phil f3ed1b3488 Merge branch 'feat/v2.13.0-backfeed' for v2.13.0 release
Build Container / build (push) Successful in 8m33s
2026-11-10 12:42:24 -07:00
PhilandClaude Opus 5.5 d400e4d838 fix: don't refetch Threads avatars whose signed URL changed (v2.13.0)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 12:21:03 -07:00
PhilandClaude Opus 5.5 55ce6a1fdd feat: backfeed of responses to crossposts (v2.13.0)
Likes, reposts, quotes and replies to a note's GoToSocial, Bluesky and
Threads copies are read back from the silos and stored as webmentions
(ADR-073): age-based schedule to 90 days, pruning of withdrawn responses,
#nobridge/#nobot opt-outs, admin backfill and Check now, Threads counts,
a Quotes group, local avatar copies, and delete tombstones.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 12:19:57 -07:00
Phil 6aaac982d5 Merge branch 'docs/v2.13.0-backfeed-adr': ADR-073 backfeed (proposed) 2026-11-10 12:04:22 -07:00
PhilandClaude Opus 5.5 2654692d15 docs: ADR-073: GoToSocial keeps threads with local posts, so backfill can reconcile
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 12:03:40 -07:00
PhilandClaude Opus 5.5 f8a429cdc1 docs: ADR-073 backfeed of responses to crossposts (proposed)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 11:59:37 -07:00
Phil afef1fc9a5 Merge branch 'fix/v2.12.1-category-unused-tags' for v2.12.1 release
Build Container / build (push) Successful in 8m3s
2026-11-10 11:28:42 -07:00
PhilandClaude Opus 5.5 b450c0c648 fix: q=category lists only tags on existing posts (v2.12.1)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 11:28:32 -07:00
Phil 13936749de Merge branch 'chore/remove-metrics': remove metrics dashboard (v2.12.0)
Build Container / build (push) Successful in 7m53s
2026-11-03 16:07:48 -07:00
PhilandClaude Opus 5.5 742e2e3ec3 Remove the metrics dashboard and monitoring package (v2.12.0)
The in-memory metrics system covered one gunicorn process at a time, was
lost on restart and was never used. Remove /admin/metrics-dashboard,
/admin/metrics, starpunk/monitoring, the HTTP/query timing, memory monitor,
feed counters, METRICS_* settings and psutil. Keep /health, /admin/health
(without its metrics block), pool stats and the feed cache. ADR-072.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 16:07:47 -07:00
Phil d6a916ba7d Merge branch 'chore/license-mit': MIT license 2026-11-03 13:43:39 -07:00
PhilandClaude Opus 5.5 e654dd84de chore: License StarPunk under MIT
MIT suits software better than CC BY (Creative Commons advises against its
licenses for code). Copyright Phil Skentelbery 2025-2026.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 13:43:39 -07:00
Phil abb75f080e Merge branch 'chore/license-cc-by': CC BY 4.0 license 2026-11-03 13:43:17 -07:00
PhilandClaude Opus 5.5 d5157a6059 chore: License StarPunk under CC BY 4.0
Replace the CC0 dedication with the Creative Commons Attribution 4.0
International license (official legal code), credited to Phil Skentelbery.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 13:43:17 -07:00
Phil 393228002b Merge branch 'chore/license-cc0': CC0 license 2026-11-03 13:42:39 -07:00
PhilandClaude Opus 5.5 874465e103 chore: Dedicate StarPunk to the public domain (CC0 1.0)
Replace the placeholder MIT notice ("Copyright (c) 2024 [Your Name]") with
the official CC0 1.0 Universal legal code.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 13:42:39 -07:00
Phil 3f5e7dc735 Merge branch 'maint/v2.11.0' for v2.11.0 release
Build Container / build (push) Successful in 7m57s
2026-11-03 13:19:44 -07:00
PhilandClaude Opus 5.5 75946be7b2 maint: Dependency upgrades and High backlog items (v2.11.0)
- All dependencies to current releases (97 known vulnerabilities -> 0),
  Python 3.14, uv pinned, requirements-test.txt, lxml/dateutil pinned
- CSRF tokens on admin forms, logout and IndieAuth consent
- IndieAuth client and crosspost link-card fetches via safe_get
- FLASK_DEBUG/FLASK_ENV default off/production; image sets them
- Foreign keys enforced on pooled connections; migration 018 removes orphans
- Rate limits shared across processes (rate_limits, migration 019)
- datetime.utcnow() and sqlite3 default adapters replaced
- compose.yaml fixed; duplicate dev_auth.py and test.ini removed

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 13:09:52 -07:00
Phil 9f17e145ff Merge branch 'fix/v2.10.2-reply-context' for v2.10.2 release
Build Container / build (push) Successful in 7m43s
2026-11-03 12:42:19 -07:00
PhilandClaude Opus 5.5 d73dd0a5b0 fix: Fetch reply contexts again (v2.10.2)
fetch_source returns (content, content_type, final_url) since v2.7.0, but
fetch_reply_context unpacked two values, so every new reply/like/repost
context was stored as an error. Parse with the final URL so relative links
resolve. Adds a test that runs the real fetch path.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 12:42:19 -07:00
Phil f85be50072 Merge branch 'docs/refresh-v2.10.1': documentation refresh 2026-11-03 12:40:19 -07:00
PhilandClaude Opus 5.5 0716984958 docs: Bring all documentation up to date for v2.10.1
- architecture/: overview, technology stack, auth (OIDC + built-in
  IndieAuth), syndication (POSSE) rewritten; new database.md,
  deployment.md and indieweb-protocols.md; legacy stack marked historical
- decisions/: reconstructed ADR-060 and ADR-063; 31 status corrections
  (superseded auth ADRs, proposed -> accepted); INDEX rebuilt
- design/: reconstructed the v1.6.0 documents the webmention code cites;
  reports for 18 v2 releases without a design folder; INDEX rebuilt
- projectplan/: BACKLOG rewritten (shipped items removed, open issues from
  the reviews added); RELEASE.md for v2.0.0-v2.10.1; INDEX
- standards/: development setup (uv, tests, DEV_MODE), testing (CI test
  stage), git and release flow, versioning, coding standards, cookies
- README, CLAUDE.md, .env.example (every setting the code reads, with
  defaults) brought up to date
- 137 broken links repaired (old absolute paths, moved files); stale ADR
  references in code comments corrected (comments only)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 12:40:19 -07:00
Phil fa551de197 Merge branch 'fix/v2.10.1-websub-hardening' for v2.10.1 release
Build Container / build (push) Successful in 7m53s
2026-11-03 11:23:36 -07:00
PhilandClaude Opus 5.5 b2043286fa fix: WebSub hardening from the adversarial review (v2.10.1)
- Per-item error handling in the worker; hard time limits on requests
- Notes with photos announced after attach; feed cache/ETag follow
  content_version
- Keep failing subscriptions until lease end; retry DNS and render errors;
  hub.mode=denied for gone topics
- Spec 5.1: decode unreserved chars, echo requested topic, ignore lease and
  secret on unsubscribe, case-insensitive Content-Type, refuse empty secret
- Limits count pending requests per callback address; callback validation;
  request_id guards the verification race (migration 017)
- Render once per topic per pass; likes only push /likes; Vary: Accept

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 11:23:36 -07:00
Phil 409b8b31b4 Merge branch 'feat/v2.10.0-websub' for v2.10.0 release
Build Container / build (push) Successful in 7m29s
2026-11-03 10:33:41 -07:00
PhilandClaude Opus 5.5 b520b8b4d3 feat: WebSub with a built-in hub (v2.10.0)
ADR-071. Topics (/, /feed.atom, /feed.rss, /feed.json, /likes, /tag/<name>)
advertise the hub and themselves in Link headers and embedded links. Note
changes bump a content version (fresh feeds in every process) and queue
deliveries. /websub accepts subscriptions for these topics only; the worker
verifies intent, delivers signed content with retries, and prunes expired
leases, all through the SSRF-safe pinned fetcher. WEBSUB_HUB_URL switches
to an external hub (publish pings). Also: the memory monitor no longer
stops after the first request; RSS self link is /feed.rss.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 10:33:41 -07:00
PhilandClaude Opus 5.5 720a1e6d6f docs: ADR-071 WebSub publishing with a built-in hub (proposed)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 10:23:00 -07:00
Phil c89c2cd5ca Merge branch 'feat/v2.9.0-micropub' for v2.9.0 release
Build Container / build (push) Successful in 7m28s
2026-11-03 10:11:23 -07:00
PhilandClaude Opus 5.5 ab6ff60f5c feat: Micropub review fixes (v2.9.0)
- mp-syndicate-to/mp-slug read from JSON properties; absent means no
  crossposting
- Sanitize rendered HTML (nh3 allowlist)
- q=source/q=category need a posting scope; draft scope creates drafts;
  private/unlisted visibility saved as drafts
- Photos that can't be fetched or found fail the request; remote photos
  via the SSRF-checked, pinned fetcher
- JSON scalars wrapped, update delete values must be arrays
- bookmark-of stored; HTML content returned as {html} (migration 015)
- published stored as UTC, bad dates 400; malformed JSON 400
- Auth: case-insensitive Bearer, header+body 400, WWW-Authenticate

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 10:11:23 -07:00
Phil 57682165a4 Merge branch 'fix/v2.8.1-migration-race' for v2.8.1 release
Build Container / build (push) Successful in 7m34s
2026-11-03 09:58:20 -07:00
PhilandClaude Opus 5.5 8c51b52ed6 fix: Hold the migration lock for the whole run (v2.8.1)
create_migrations_table() committed right after BEGIN IMMEDIATE, and
executescript() commits before running, so gunicorn workers applied
migrations concurrently; the v2.8.0 deploy failed to boot once with
'duplicate column name'. Run migration statements with conn.execute inside
the transaction and don't commit when creating the tracking table.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 09:58:20 -07:00
Phil 42c77e34b7 Merge branch 'feat/v2.8.0-microformats' for v2.8.0 release
Build Container / build (push) Successful in 9m35s
2026-11-03 09:36:53 -07:00
PhilandClaude Opus 5.5 438465592c feat: Microformats2 review fixes (v2.8.0)
- Responses as u-comment/u-like/u-repost/u-bookmark/u-mention h-cites
  with author h-cards; source's own url and published (migration 014)
- dt-* and Micropub published in RFC 3339 UTC
- Title only for a real Markdown heading (hashtag notes stay notes)
- Ignore webmentions from a note's own syndicated copies
- Micropub/IndieAuth links independent of Webmention; rel=me/rel=author
- Representative h-card and authorship algorithm (starpunk/mf2_utils.py)
- Tag/likes h-feeds get url and author; previews get u-uid

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 09:36:49 -07:00
Phil c8a7ef309f Merge branch 'feat/v2.7.2-webmention-shoulds' for v2.7.2 release
Build Container / build (push) Successful in 9m28s
2026-11-03 09:01:42 -07:00
PhilandClaude Opus 5.5 0640730416 feat: Pin webmention connections; re-send when responses change (v2.7.2)
- Connect to the address that passed the SSRF check (Host header and SNI
  carry the hostname, so TLS still verifies it), closing the DNS rebinding
  gap (spec 4.5). The webmention POST uses the same path.
- Treat a change in the responses a note shows as an update and re-send its
  webmentions (spec 3.1.4), from validation and from admin moderation.
  Unchanged re-validations send nothing, so there is no ping-pong.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 09:01:38 -07:00
Phil 51f711ea4a Merge branch 'ci/v2.7.1-pytest' for v2.7.1 release
Build Container / build (push) Successful in 9m13s
2026-11-03 08:38:40 -07:00
PhilandClaude Opus 5.5 6a76ee26af ci: Run the test suite before building release images (v2.7.1)
Add a 'test' stage to the Containerfile (production base image plus pytest)
and build it in the release workflow before the push step, so a failing test
stops the release.

Fix the eight tests that were failing: media validation returns a PIL Image
since v1.5.0, and the migration test needs TESTING to skip production config
checks.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 08:28:08 -07:00
Phil 43ec7cb8be Merge branch 'feat/v2.7.0-webmention-spec' for v2.7.0 release
Build Container / build (push) Successful in 11s
2026-11-03 08:07:12 -07:00
PhilandClaude Opus 5.5 4aaa7a8122 feat: W3C Webmention compliance (v2.7.0)
Sending: RFC 8288 Link header parsing, rel token matching and document
order for <link>/<a>, endpoint re-discovery on every send, re-send to
previous targets on update/delete/undelete, 410 Gone for deleted notes,
no redirect following on the POST, 429 retried.

Receiving: exact target match per media type, Accept header, final URL
for relative links, only published notes accepted, fragments ignored,
moderation kept on updates, transient errors keep valid mentions.

Security: every redirect hop SSRF-checked, all resolved addresses
checked, broader blocklist, total time and size limits.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 08:07:04 -07:00
Phil 0cfc9965df Merge branch 'fix/v2.6.9-webmention-processing' for v2.6.9 release
Build Container / build (push) Successful in 12s
2026-11-03 07:53:19 -07:00
PhilandClaude Opus 5.5 106495d114 fix: Webmention processing works end to end (v2.6.9)
- Normalise mf2 author values: u-photo with alt text is a dict in mf2py and
  broke the database update for most real senders
- Pass str(response.url) to urljoin during endpoint discovery; the TypeError
  marked every target as having no endpoint. Retry on discovery errors
- Hold a cross-process flock so only one gunicorn process runs the worker
- Keep author URL/photo only if http(s) (XSS via javascript: URLs)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 07:53:19 -07:00
Phil 3c1e109358 Merge branch 'fix/v2.6.8-webmention-worker' for v2.6.8 release
Build Container / build (push) Successful in 20s
2026-11-03 07:41:11 -07:00
PhilandClaude Opus 5.5 d43361d505 fix: Webmention worker runs inside the app context (v2.6.8)
The worker thread called the SSRF checks, which read current_app.config,
without an app context, so every incoming and outgoing webmention failed with
'Working outside of application context'. A teardown_appcontext hook also
stopped the worker whenever a request ended.

Pass the app to WebmentionWorker and run each poll in app.app_context(), like
SyndicationWorker, and drop the teardown hook.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-03 07:41:06 -07:00
Phil f924413934 Merge branch 'fix/v2.6.7-admin-layout' for v2.6.7 release
Build Container / build (push) Successful in 7s
2026-11-02 18:24:15 -07:00
PhilandClaude Opus 5.5 4b2ae4c7fc fix: Admin pages line up with the header (v2.6.7)
Every page sat in the 40rem reading column, so the admin wrapper's 60rem
never applied and the notes table, which can't shrink that far, ran past
the header's right edge. Admin pages now set a 60rem frame for header,
content and footer; post previews wrap, and tables scroll sideways on
narrow screens.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-02 18:24:14 -07:00
Phil d3148063e3 Merge branch 'fix/v2.6.6-inline-image-crosspost' for v2.6.6 release
Build Container / build (push) Successful in 15s
2026-11-02 18:12:51 -07:00
PhilandClaude Opus 5.5 39c8259f3f fix: Crossposts include photos embedded in the post text (v2.6.6)
Some Micropub clients (OutPost) upload a photo to the media endpoint and
then embed it in content instead of sending it as `photo`. Those posts
have no attached media, so crossposts went out without the image.

Images in the text that are in this site's media library are now sent
like attached photos, with their alt text, and GoToSocial's text no
longer carries the image Markdown.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-02 18:11:27 -07:00
Phil 1ce88b38a5 Merge branch 'fix/v2.6.5-micropub-media' for v2.6.5 release
Build Container / build (push) Successful in 6s
2026-11-01 10:53:37 -07:00
PhilandClaude Opus 5.5 4ffe36f9d3 fix: Photos without alt text use the post's text (v2.6.5)
Uncaptioned photos rendered alt="Image", which mf2 parsers report as
the photo's alt text. They now take the post's text (excerpt, up to 150
characters), falling back to "Image" only for text-free posts.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-01 10:44:54 -07:00
PhilandClaude Opus 5.5 a551770c76 fix: Replies and photo posts read as the right post type (v2.6.5)
Checked rendered pages with mf2py against W3C Post Type Discovery:

- Note pages put u-in-reply-to inside a bare h-cite, making it the
  h-cite's property, so replies parsed as notes (or photo posts). The
  reply context is now the entry's u-in-reply-to h-cite.
- Home page previews had no reply markup; they now show "In reply to"
  with u-in-reply-to.
- Previews always emit e-content, so a text-free photo post can't get
  an implied name.
- Text-free page titles follow PTD order (Repost, Like, Reply, Photo).

tests/test_post_types.py runs PTD on parsed note pages and the home page.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-01 10:06:44 -07:00
PhilandClaude Opus 5.5 d0bfc80c2e fix: Micropub photo posts and failed creates (v2.6.5)
- A create that failed after the note was written (e.g. the same photo
  twice) answered 500 but left the post published without photos.
  Photos are now resolved first, and a failed note is removed.
- Pooled connections roll back unfinished writes, so one failure no
  longer locks the database for the next ~10 s.
- The same photo twice in one post is shown once.
- Photo posts no longer need text (Micropub doesn't require content,
  and q=config already advertised a photo-only post type). Updates may
  clear the text while photos remain. Crossposts start with the link;
  the page is titled "Photo".

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-01 09:50:30 -07:00
PhilandClaude Opus 5.5 d16463ef8a Merge branch 'fix/v2.6.4-tag-order' for v2.6.4 release
Build Container / build (push) Successful in 7s
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-01 08:59:19 -07:00
PhilandClaude Opus 5.5 4b56f7cd4b fix: Tags keep the order they were entered in (v2.6.4)
add_tags_to_note() rewrites note_tags in entry order on every save, so
note_tags.id already records it; get_note_tags(entered_order=True) reads
it back. Crossposts (Threads topic, which hashtags fit on Bluesky), the
editor and Micropub q=source/update use it, so saving no longer
reorders tags alphabetically. Public pages stay alphabetical.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-01 08:59:18 -07:00
PhilandClaude Opus 5.5 2474628646 Merge branch 'fix/v2.6.3-link-cards' for v2.6.3 release
Build Container / build (push) Successful in 8s
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-01 08:12:25 -07:00
PhilandClaude Opus 5.5 0fe341161e feat: Crosspost cards show the note's own link (v2.6.3)
When a note links somewhere (an article, say), the first link gets the
preview card instead of the note (ADR-070). Bluesky cards are built from
the page's OG/Twitter/HTML title, description and image (thumb shrunk
under 1 MB); Threads previews the first link in the text. Markdown links
no longer lose their URL on Bluesky (link text becomes a facet) or Threads
("text (url)"); Threads keeps to Meta's 5-link limit. Page fetches are
public-address only, checked at every redirect, size-capped, and never
fail the post.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-01 08:12:24 -07:00
PhilandClaude Opus 5.5 a62077ab30 Merge branch 'fix/v2.6.2-threads-link-preview' for v2.6.2 release
Build Container / build (push) Successful in 7s
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 21:38:00 -06:00
PhilandClaude Opus 5.5 fade75fd2d fix: Let Threads build the link preview itself (v2.6.2)
An explicit link_attachment for a URL Meta hasn't fetched yet fails at
threads_publish with "Invalid Link Attachment" (subcode 4279047), which
is every new note; it only went out on the retry. Tested live with fresh
URLs: with link_attachment it fails, without it it publishes first time
and Threads still sets link_attachment_url from the first link in the
text. Reposts likewise.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 21:38:00 -06:00
PhilandClaude Opus 5.5 0dd59fbc18 Merge branch 'fix/v2.6.1-threads-publish-retry' for v2.6.1 release
Build Container / build (push) Successful in 6s
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 21:18:11 -06:00
PhilandClaude Opus 5.5 11e09a0d02 fix: Retry Threads publish errors, show full error detail (v2.6.1)
Meta's threads_publish returned a one-off HTTP 400 "Fatal" for a plain
text post; the same post published fine on retry. Publish-step errors
are now retryable unless they are auth/permission errors (codes 10, 190,
200); the retry's existing-post check prevents doubles. Errors carry the
Graph code, subcode, user message and fbtrace id. A short pause before
the first container status poll.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 21:18:11 -06:00
PhilandClaude Opus 5.5 565adaeecf Merge branch 'feature/v2.6.0-threads' for v2.6.0 release
Build Container / build (push) Successful in 7s
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 20:54:15 -06:00
PhilandClaude Opus 5.5 53d10ed271 feat: Crosspost notes to Threads (v2.6.0)
Third crossposting target on the ADR-065 loop (ADR-069). Text posts get a
link preview, 1 image posts as IMAGE and 2-4 as a CAROUSEL (Meta fetches
them from /media/), with alt text and the first tag as topic_tag. Posts are
shortened to 500 characters, counting non-ASCII as UTF-8 bytes. Edits don't
follow (the API has no edit); unpublish and delete remove the post. Retried
creates check for an existing post first.

The 60-day token is refreshed daily from the worker loop and kept in
DATA_PATH/threads_token.json under a file lock; a new configured token
replaces it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 20:46:56 -06:00
PhilandClaude Opus 5.5 e8c1b03980 feat: Likes and reposts (v2.5.0)
Build Container / build (push) Successful in 6s
Micropub like-of/repost-of with optional comment; likes on /likes, off the
home page and feeds; reposts in the stream. Native favourites/boosts on
GoToSocial and likes/reposts/quotes on Bluesky when the target is on that
network, link posts for other reposts. Fix double slash in outgoing
webmention source URLs. ADR-068.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 20:16:21 -06:00
PhilandClaude Opus 5.5 0d3a5cd519 feat: Micropub update, delete/undelete, drafts, replies (v2.4.0)
Build Container / build (push) Successful in 6s
Update (replace/add/delete) for content, name, category, photo,
in-reply-to, post-status; soft delete with undelete; post-status drafts;
in-reply-to; multipart and remote photos; q=source lists and properties[];
q=category; media scope. Fix media endpoint 404 at the advertised
/micropub/media path. ADR-067.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 19:53:18 -06:00
Phil 88088e10e1 Merge branch 'fix/v2.3.3-micropub-title-source-url' for v2.3.3 release
Build Container / build (push) Successful in 7s
2026-10-31 19:34:24 -06:00
PhilandClaude Opus 5.5 391bdb9db3 fix: Micropub titles, q=source URL, single title display (v2.3.3)
Micropub `name` was dropped; store it as the note's '# Heading' first
line, which is how StarPunk titles work. q=source returned /notes/<slug>
(a 404) and a name for every note; it now returns the real permalink and
name only for titled notes, without the heading in content. Titled notes
no longer show their title twice (Note.body_html, excerpt).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 19:34:23 -06:00
PhilandClaude Opus 5.5 c51cf726b0 fix: Header shows Admin when logged in (v2.3.2)
Build Container / build (push) Successful in 9s
Public pages never checked the session, so they always showed Log in.
Set g.me in before_request when a session cookie is present.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 18:57:12 -06:00
PhilandClaude Opus 5.5 2616ef5060 fix: Put Log in back in the header nav (v2.3.1)
Build Container / build (push) Successful in 7s
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 18:46:33 -06:00
Phil 0ae16bd1fc Merge branch 'feature/v2.3.0-style-pass' for v2.3.0 release
Build Container / build (push) Successful in 8s
2026-10-31 18:43:20 -06:00
PhilandClaude Opus 5.5 cc09098741 style: Public page pass (v2.3.0)
Fix note metadata inheriting page-footer styles, show full notes instead of
HTML cut at 300 chars, auto-link bare URLs, one-line meta, SITE_NAME in the
header and titles, serif reading font, dark mode.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 18:24:29 -06:00
Phil 16925663a3 Merge branch 'feature/v2.2.0-bluesky' for v2.2.0 release
Build Container / build (push) Successful in 7s
2026-10-31 14:19:17 -06:00
PhilandClaude Opus 5.5 e92c27a84c feat: Crosspost notes to Bluesky (v2.2.0)
BlueskyClient behind the ADR-065 reconcile loop: plain text with link and
hashtag facets, images under 1 MB with alt text, link cards, app-password
sessions with refresh. Deletes follow; edits don't (ADR-066).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 14:19:17 -06:00
PhilandClaude Opus 5.5 ed63e70c46 Merge branch 'feature/v2.1.0-gotosocial-syndication' for v2.1.0 release
Build Container / build (push) Successful in 8s
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 12:46:28 -06:00
PhilandClaude Opus 5.5 57b02ad6c7 feat: Crosspost notes to GoToSocial (v2.1.0)
Published notes are copied to a GoToSocial (Mastodon-API) account and
edits, unpublishing and deletes follow them (ADR-065). On by default for
new notes, untickable in the editor and selectable from Micropub via
mp-syndicate-to. A background worker reconciles each (note, target)
row with atomic claims, an idempotency key and a version so edits during
a send are not lost. Synced copies are linked with u-syndication.

Also fixes the Micropub Location header and search result links, which
pointed at /notes/<slug> (404), and the nested delete form on the edit
page.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 12:46:26 -06:00
PhilandClaude Opus 5.5 94e8c672df release: v2.0.2 (CI registry login)
Build Container / build (push) Successful in 18s
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 12:20:01 -06:00
PhilandClaude Opus 5.5 011f7cc87d ci: Log in to the registry with a write:package token
The automatic Actions token is refused by the container registry
(run 107: unauthorized). Use the REGISTRY_TOKEN repository secret.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 12:13:43 -06:00
PhilandClaude Opus 5.5 f46130569f Merge branch 'fix/v2.0.1-ci-and-feed-urls' for v2.0.1 release
Build Container / build (push) Failing after 12s
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 12:12:29 -06:00
PhilandClaude Opus 5.5 b4e6cf3cc3 fix: Build absolute URLs from SITE_URL, not the request (v2.0.1)
Behind Caddy the app is reached over plain HTTP, so url_for(_external=True)
produced http:// links on the https site: the RSS/OPML <link>s and the
microformats u-url/u-uid permalinks that IndieWeb parsers treat as
canonical. Templates now use absolute_url_for(), shared with the
IndieAuth metadata.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 12:12:28 -06:00
PhilandClaude Opus 5.5 2d0568d9c2 Build the container image with the server's rootless BuildKit
The runner no longer has a Docker daemon (it gave every workflow root on
the server), and this workflow never worked anyway: it logged in to
gitea:3000, which the daemon couldn't resolve. Log in to
git.thesatelliteoflove.com and build through the BuildKit socket instead.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 11:53:59 -06:00
PhilandClaude Opus 5.5 f117913797 Merge branch 'feature/v2.0.0-oidc' for v2.0.0 release
Build Container / build (push) Failing after 2m38s
Includes the v1.6.0-rc.2 source recovered from production.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 11:45:20 -06:00
PhilandClaude Opus 5.5 b84cf23e1a feat!: OIDC admin login and built-in IndieAuth server (v2.0.0)
Admin login now uses OpenID Connect (PKCE, nonce, ID token verified
against the provider JWKS) and only OIDC_ADMIN_SUB may log in. Micropub
tokens are issued and verified by StarPunk itself, with a consent screen
behind the OIDC session and a Connected Apps page to revoke them, so an
external IndieAuth server (Gondulf) is no longer needed.

Existing notes, media, tags, webmentions, author profile and sessions
are unchanged. Migration 011 adds new tables only.

BREAKING CHANGE: production requires OIDC_ISSUER, OIDC_CLIENT_ID,
OIDC_CLIENT_SECRET and OIDC_ADMIN_SUB. The ADMIN_ME page must point
Micropub apps at StarPunk's IndieAuth endpoints. See ADR-064.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 11:38:27 -06:00
PhilandClaude Opus 5.5 d3fa5b7a2a chore: Recover v1.6.0-rc.2 source from production image
The v1.6.0 work (webmentions, reply contexts, rate limiter, migration
010) was built and deployed as git.thesatelliteoflove.com/phil/starpunk:v1.6.0-rc.2
but never committed. This restores the application code from /app in
the running container so later work builds on what production runs.
Tests and docs for v1.6.0 were not in the image and are not recovered.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-31 11:17:51 -06:00
phil 06a8aabc01 Merge branch 'feature/v1.5.0' for v1.5.0 release
Build Container / build (push) Failing after 11s
2025-12-17 15:58:32 -07:00
philandClaude 4ee2c189ae release: v1.5.0 - Quality of Life Improvements
IndieAuth Authentication:
- Corrected W3C IndieAuth specification compliance
- Uses response_type=id for authentication-only flow per spec
- Discovers endpoints from user profile URL
- Removed hardcoded indielogin.com service
- DEPRECATED: INDIELOGIN_URL config (now auto-discovered)

Timestamp-Based Slugs (ADR-062):
- Default slugs now use YYYYMMDDHHMMSS format
- Unique collision handling with numeric suffix

Debug File Management:
- Controlled by DEBUG_SAVE_FAILED_UPLOADS config
- Auto-cleanup of files older than 7 days
- 100MB disk space protection
- Filename sanitization for security

Performance:
- N+1 query fix in feed generation
- Batch media loading for feed notes

Data Integrity:
- Atomic variant generation with temp files
- Database/filesystem consistency on failure

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-17 15:57:45 -07:00
philandClaude c94cb377d3 release: Bump version to 1.5.0-rc.1 for testing
Release candidate for v1.5.0 with IndieAuth hotfix.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-17 14:25:33 -07:00
philandClaude 2bd971f3d6 fix(auth): Implement IndieAuth endpoint discovery per W3C spec
BREAKING: Removes INDIELOGIN_URL config - endpoints are now properly
discovered from user's profile URL as required by W3C IndieAuth spec.

- auth.py: Uses discover_endpoints() to find authorization_endpoint
- config.py: Deprecation warning for obsolete INDIELOGIN_URL setting
- auth_external.py: Relaxed validation (allows auth-only flows)
- tests: Updated to mock endpoint discovery

This fixes a regression where admin login was hardcoded to use
indielogin.com instead of respecting the user's declared endpoints.

Version: 1.5.0-hotfix.1

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-17 13:52:36 -07:00
philandClaude 84e693fe57 release: Bump version to 1.5.0
🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-17 12:36:16 -07:00
philandClaude 975046abc7 test: Expand coverage to 90% for v1.5.0
Added 3 MPO format tests per v1.5.0 Phase 5 requirements:
- test_mpo_detection_and_conversion: Verify MPO->JPEG conversion
- test_mpo_dimensions_preserved: Verify dimensions maintained
- test_mpo_full_upload_flow: Test complete upload workflow

MPO (Multi-Picture Object) format handling was implemented in v1.4.2
but was previously untested. These tests ensure the format detection
and JPEG conversion work correctly for iPhone portrait/depth photos.

Test count: 924 -> 927 tests (all passing)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-17 11:56:03 -07:00
philandClaude 21fa7acfbb feat(media): Make variant generation atomic with database
Per v1.5.0 Phase 4:
- Generate variants to temp directory first
- Perform database inserts in transaction
- Move files to final location before commit
- Clean up temp files on any failure
- Add startup recovery for orphaned temp files
- All media operations now fully atomic

Changes:
- Modified generate_all_variants() to return file moves
- Modified save_media() to handle full atomic operation
- Add cleanup_orphaned_temp_files() for startup recovery
- Added 4 new tests for atomic behavior
- Fixed HEIC variant format detection
- Updated variant failure test for atomic behavior

Fixes:
- No orphaned files on database failures
- No orphaned DB records on file failures
- Startup recovery detects and cleans orphans

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-17 11:26:26 -07:00
philandClaude b689e02e64 perf(feed): Batch load media and tags to fix N+1 query
Per v1.5.0 Phase 3: Fix N+1 query pattern in feed generation.

Implementation:
- Add get_media_for_notes() to starpunk/media.py for batch media loading
- Add get_tags_for_notes() to starpunk/tags.py for batch tag loading
- Update _get_cached_notes() in starpunk/routes/public.py to use batch loading
- Add comprehensive tests in tests/test_batch_loading.py

Performance improvement:
- Before: O(n) queries (1 query per note for media + 1 query per note for tags)
- After: O(1) queries (2 queries total: 1 for all media, 1 for all tags)
- Maintains same API behavior and output format

All tests passing: 920 passed in 360.79s

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-17 10:42:44 -07:00
philandClaude 1b45a64920 feat: v1.5.0 Phase 2 - Debug File Management
Implement debug file management system with configuration controls,
automatic cleanup, and security improvements per v1.5.0 Phase 2.

## Changes

### Configuration (config.py)
- Add DEBUG_SAVE_FAILED_UPLOADS (default: false, production-safe)
- Add DEBUG_FILE_MAX_AGE_DAYS (default: 7 days)
- Add DEBUG_FILE_MAX_SIZE_MB (default: 100MB)

### Media Validation (media.py)
- Check config before saving debug files
- Sanitize filenames to prevent path traversal
- Pattern: alphanumeric + "._-", truncated to 50 chars
- Add cleanup_old_debug_files() function
  * Age-based cleanup (delete files older than MAX_AGE)
  * Size-based cleanup (delete oldest if total > MAX_SIZE)

### Application Startup (__init__.py)
- Run cleanup_old_debug_files() on startup
- Automatic maintenance of debug directory

### Tests (test_debug_file_management.py)
- 15 comprehensive tests
- Config defaults and overrides
- Debug file saving behavior
- Filename sanitization security
- Cleanup age and size limits
- Startup integration

## Security Improvements
- Debug saving disabled by default (production-safe)
- Filename sanitization prevents path traversal
- Automatic cleanup prevents disk exhaustion

## Acceptance Criteria
- [x] Configuration options added
- [x] Debug saving disabled by default
- [x] Filename sanitized before saving
- [x] Cleanup runs on startup
- [x] Old files deleted based on age
- [x] Size limit enforced

All tests pass. Ready for architect review.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-17 10:05:42 -07:00
philandClaude 3f1f82a749 feat(slugs): Implement timestamp-based slugs per ADR-062
Replaces content-based slug generation with timestamp format YYYYMMDDHHMMSS.
Simplifies slug generation and improves privacy by not exposing note content in URLs.

Changes:
- Add generate_timestamp_slug() to slug_utils.py
- Update notes.py to use timestamp slugs for default generation
- Sequential collision suffix (-1, -2) instead of random
- Custom slugs via mp-slug continue to work unchanged
- 892 tests passing (+18 new timestamp slug tests)

Per ADR-062 and v1.5.0 Phase 1 specification.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-17 09:49:30 -07:00
philandClaude 92e7bdd342 feat(tests): Phase 0 - Fix flaky and broken tests
Implements Phase 0 of v1.5.0 per ADR-012 and RELEASE.md.

Changes:
- Remove 5 broken multiprocessing tests (TestConcurrentExecution, TestPerformance)
- Fix brittle XML assertion tests (check semantics not quote style)
- Fix test_debug_level_for_early_retries logger configuration
- Rename test_feed_route_streaming to test_feed_route_caching (correct name)

Results:
- Test count: 879 → 874 (5 removed as planned)
- All tests pass consistently (verified across 3 runs)
- No flakiness detected

References:
- ADR-012: Flaky Test Removal and Test Quality Standards
- docs/projectplan/v1.5.0/RELEASE.md Phase 0

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-17 09:24:12 -07:00
philandClaude 0acefa4670 docs: Update Phase 0 with specific test fix requirements
Per ADR-012, Phase 0 now specifies:
- 5 tests to REMOVE (broken multiprocessing)
- 4 tests to FIX (brittle assertions)
- 1 test to RENAME (misleading name)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 20:45:41 -07:00
philandClaude 9dcc5c5710 docs: v1.5.0 planning - ADR-062, release plan, and design docs
- ADR-062: Timestamp-based slug format (supersedes ADR-007)
- Updated v1.5.0 RELEASE.md with 6-phase plan
- Updated BACKLOG.md with deferred N+1 query locations
- Developer questions and architect responses for Phase 1

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 19:38:01 -07:00
philandClaude 7be2fb0f62 docs: Add v1.5.0 "Trigger" release definition
Focus: Cleanup, test coverage, and quality of life improvements

Goals:
- 90% test coverage target
- MPO format test coverage (High backlog item)
- Debug file storage cleanup (Medium backlog item)
- Filename sanitization in debug path (Medium backlog item)
- N+1 query pattern fix (Medium backlog item)
- Atomic variant generation (Medium backlog item)
- Default slug change to timestamp format (Medium backlog item)

Backlog items marked as scheduled for v1.5.0.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 19:04:43 -07:00
philandClaude 730eb8d58c docs: Add v1.4.2 review documents and update backlog
Build Container / build (push) Failing after 15s
- Fix stale docstring in media.py (4096 -> 12000)
- Add developer review document
- Add architect review document
- Update backlog with identified issues

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 18:57:39 -07:00
philandClaude 6682339a86 fix(media): Increase max dimension to 12000px for modern phone cameras
Modern iPhones (48MP) and other phones produce images larger than 4096px.
Since optimize_image() resizes them anyway, the input limit was too
restrictive. Increased from 4096x4096 to 12000x12000.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 18:29:21 -07:00
philandClaude d416463242 fix(media): Add MPO format support for iPhone portrait photos
iPhones use MPO (Multi-Picture Object) format for depth/portrait photos.
This format contains multiple JPEG images (main + depth map). Pillow
opens these as MPO format which wasn't in our allowed types.

Added automatic MPO to JPEG conversion by extracting the primary image.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 18:23:23 -07:00
philandClaude 25b8cbd79d chore: Add format detection logging for debugging
Logs the detected image format when a file is rejected to help
diagnose why iPhone uploads are being rejected.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 18:14:05 -07:00
philandClaude 042505d5a6 fix(media): Add MAX_CONTENT_LENGTH and debug file capture - v1.4.2
- Set Flask MAX_CONTENT_LENGTH to 50MB (matches MAX_FILE_SIZE)
- Save failed uploads to data/debug/ for analysis
- Log magic bytes when both Pillow and HEIC parsers fail

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 18:08:36 -07:00
philandClaude 72f3d4a55e fix(media): Save failed uploads to debug/ for analysis - v1.4.2
When an image fails both Pillow and HEIC parsing, save the file
to data/debug/ for manual analysis.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 18:06:51 -07:00
philandClaude e8ff0a0dcb fix(media): Add debug logging for unrecognized image formats - v1.4.2
Logs magic bytes when both Pillow and HEIC parsers fail,
to help diagnose what format the file actually is.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 18:06:09 -07:00
philandClaude 9bc6780a8e fix(media): Handle HEIC files with wrong extension - v1.4.2
iOS sometimes saves HEIC with .jpeg extension. Pillow fails to open
these as JPEG, so now we fallback to trying pillow-heif directly
when initial open fails.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 17:54:50 -07:00
philandClaude e4e481d7cf feat(media): Add HEIC/HEIF image support - v1.4.2
- Add pillow-heif dependency for iPhone photo support
- Auto-convert HEIC to JPEG (browsers can't display HEIC)
- Graceful error if pillow-heif not installed
- Handles RGBA/P mode conversion to RGB

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 17:45:53 -07:00
philandClaude 07f351fef7 feat(media): Add comprehensive logging for media uploads - v1.4.1
Implements media upload logging per docs/design/v1.4.1/media-logging-design.md

Changes:
- Add logging to save_media() in starpunk/media.py:
  * INFO: Successful uploads with file details
  * WARNING: Validation/optimization/variant failures
  * ERROR: Unexpected system errors
- Remove duplicate logging in Micropub media endpoint
- Add 5 comprehensive logging tests in TestMediaLogging class
- Bump version to 1.4.1
- Update CHANGELOG.md

All media upload operations now logged for debugging and observability.
Validation errors, optimization failures, and variant generation issues
are tracked at appropriate log levels. Original functionality unchanged.

Test results: 28/28 media tests pass, 5 new logging tests pass

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 17:22:22 -07:00
philandClaude fd92a1d1eb fix: Clean up variant files in delete_media() - v1.4.0
Following design approved by architect in docs/design/v1.4.0/

Changes to starpunk/media.py:
- Update delete_media() to fetch and delete variant files from disk
- Query media_variants table before deletion for file paths
- Use best-effort cleanup with try/except for each file
- Log individual file deletion failures as warnings
- Update final log to show total files deleted (original + variants)
- Update module docstring to reflect v1.4.0 capabilities:
  * 50MB max upload, 10MB max output
  * Image variants (thumb, small, medium, large)
  * Tiered resize strategy

This fixes the issue where variant files were left orphaned on disk
when media was deleted. The database CASCADE already deleted variant
records, but the physical files remained.

All tests pass: uv run pytest tests/test_media_upload.py -v (23/23)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 11:10:21 -07:00
philandClaude 68d1a1d879 feat: v1.4.0 Phase 5 - Testing & Documentation
Complete Phase 5 of v1.4.0 "Media" release with comprehensive release
preparation.

CHANGELOG Updates:
- Added v1.4.0rc1 section with comprehensive release notes
- Documented all four phases:
  * Large Image Support (50MB limit, tiered resize)
  * Image Variants (thumb, small, medium, large)
  * Micropub Media Endpoint (W3C compliant)
  * Enhanced Feed Media (Media RSS, JSON Feed extension)
- Standards compliance documented (W3C Micropub, Media RSS 2.0)
- Storage impact and backwards compatibility notes
- Configuration options (STARPUNK_ABOUT_URL)

Version Bump:
- Updated starpunk/__init__.py to 1.4.0rc1
- Version info tuple includes rc1 designator

Implementation Report:
- Created docs/design/v1.4.0/2025-12-16-v140-implementation-complete.md
- Complete summary of all five phases
- Test results: 850 passing, 19 failing
- Known issues documented (flaky tests, feed cache)
- Acceptance criteria checklist
- Deployment and rollback procedures
- Release recommendation: Ready for RC testing

Test Results:
- Full test suite executed: 850 passed, 19 failed (98% pass rate)
- Core v1.4.0 functionality verified working
- Failures in known flaky tests and unrelated areas
- 5 minutes 58 seconds execution time

Release Status:
- All five phases complete
- Ready for release candidate testing
- Manual verification recommended before final release

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 08:18:54 -07:00
philandClaude 00f21d2a51 docs: Add Phase 4 implementation report
Document completed Phase 4 implementation with test results,
verification output, and backwards compatibility notes.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 08:01:06 -07:00
philandClaude 83dc488457 feat: v1.4.0 Phase 4 - Enhanced Feed Media
Implement Phase 4 of v1.4.0 Media release - Enhanced Feed Media support.

RSS Feed Enhancements (starpunk/feeds/rss.py):
- Wrap size variants in <media:group> elements
- Add <media:content> for large/medium/small variants with attributes:
  url, type, medium, isDefault, width, height, fileSize
- Add <media:thumbnail> for thumb variant with dimensions
- Add <media:title type="plain"> for image captions
- Implement isDefault logic: largest available variant (large→medium→small fallback)
- Maintain backwards compatibility for media without variants (legacy fallback)

JSON Feed Enhancements (starpunk/feeds/json_feed.py):
- Add _starpunk.about URL (configurable via STARPUNK_ABOUT_URL config)
- Add _starpunk.media_variants array with variant data when variants exist
- Each variant entry includes: url, width, height, size_in_bytes, mime_type

ATOM Feed Enhancements (starpunk/feeds/atom.py):
- Add title attribute to enclosure links for captions
- Keep simple (no variants in ATOM per design decision)

Test Updates (tests/test_feeds_rss.py):
- Update streaming media test to search descendants for media:content
- Now inside media:group for images with variants (v1.4.0 behavior)

Per design document: /docs/design/v1.4.0/media-implementation-design.md
Following ADR-059: Full Feed Media Standardization

Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-16 07:47:56 -07:00
philandClaude c64feaea23 feat: v1.4.0 Phase 3 - Micropub Media Endpoint
Implement W3C Micropub media endpoint for external client uploads.

Changes:
- Add POST /micropub/media endpoint in routes/micropub.py
  - Accept multipart/form-data with 'file' field
  - Require bearer token with 'create' scope
  - Return 201 Created with Location header
  - Validate, optimize, and generate variants via save_media()

- Update q=config response to advertise media-endpoint
  - Include media-endpoint URL in config response
  - Add 'photo' post-type to supported types

- Add photo property support to Micropub create
  - extract_photos() function to parse photo property
  - Handles both simple URL strings and structured objects with alt text
  - _attach_photos_to_note() function to attach photos by URL
  - Only attach photos from our server (by URL match)
  - External URLs logged but ignored (no download)
  - Maximum 4 photos per note (per ADR-057)

- SITE_URL normalization pattern
  - Use .rstrip('/') for consistent URL comparison
  - Applied in media endpoint and photo attachment

Per design document: docs/design/v1.4.0/media-implementation-design.md

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 18:32:21 -07:00
philandClaude 501a711050 feat(media): Add image variant support - v1.4.0 Phase 2
Implement automatic generation of multiple image sizes for responsive
delivery and enhanced feed optimization.

Changes:
- Add migration 009 for media_variants table with CASCADE delete
- Define variant specs: thumb (150x150 crop), small (320px), medium (640px), large (1280px)
- Implement generate_variant() with center crop for thumbnails and aspect-preserving resize
- Implement generate_all_variants() with try/except cleanup, pass year/month/optimized_bytes explicitly
- Update save_media() to generate all variants after saving original
- Update get_note_media() to include variants dict (backwards compatible - only when variants exist)
- Record original as 'original' variant type

Technical details:
- Variants use explicit year/month parameters to avoid fragile path traversal
- Pass optimized_bytes to avoid redundant file I/O
- File cleanup on variant generation failure
- Skip generating variants larger than source image
- variants key only added to response when variants exist (pre-v1.4.0 compatibility)

All existing tests pass. Phase 2 complete.

Per design document: /home/phil/Projects/starpunk/docs/design/v1.4.0/media-implementation-design.md

Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 18:19:24 -07:00
philandClaude 1b51c82656 feat: Implement v1.4.0 Phase 1 - Large Image Support
Implement tiered resize strategy for large images per v1.4.0 design:

Changes:
- Increase MAX_FILE_SIZE from 10MB to 50MB
- Add MAX_OUTPUT_SIZE constant (10MB target after optimization)
- Add MIN_QUALITY and MIN_DIMENSION constants
- Add get_optimization_params() for tiered strategy:
  - <=10MB: 2048px max, 95% quality
  - 10-25MB: 1600px max, 90% quality
  - 25-50MB: 1280px max, 85% quality
- Update optimize_image() signature to return 4-tuple (img, w, h, bytes)
- Implement iterative quality reduction if output >10MB
- Add animated GIF detection and size check in validate_image()
- Update save_media() to use new optimize_image() return value
- Fix GIF format preservation during optimization
- Update tests to match new optimize_image() signature

All existing tests pass. Ready for Phase 2 (Image Variants).

Following design in:
/home/phil/Projects/starpunk/docs/design/v1.4.0/media-implementation-design.md

Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 17:57:44 -07:00
philandClaude 5ea9c8f330 release: Bump version to 1.3.1
Build Container / build (push) Failing after 13s
Promote v1.3.1-rc.1 to stable release "Syndicate Tags"

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 14:36:59 -07:00
philandClaude 98692c35db Merge branch 'feature/feed-tags' for v1.3.1
Add tags/categories to RSS, Atom, and JSON Feed formats.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 13:28:16 -07:00
philandClaude 61cba2fa6d release: Bump version to 1.3.1rc1
Prepare release candidate for v1.3.1 "Syndicate Tags"

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 13:27:55 -07:00
philandClaude cba24ab06f fix: Use _cached_tags for Note model compatibility in feeds
The Note model has a `tags` property that uses `_cached_tags` internally.
Setting `tags` directly via object.__setattr__ conflicts with the property.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 13:25:28 -07:00
philandClaude 9b26de7b05 feat: Add tag/category support to all feed formats (v1.3.1)
Implements tag/category rendering in RSS 2.0, Atom 1.0, and JSON Feed 1.1
syndication feeds. Tags added in v1.3.0 are now visible to feed readers.

Changes:
- Load tags in feed generation (_get_cached_notes)
- Add <category> elements to RSS 2.0 feeds
- Add <category term/label> to Atom 1.0 feeds
- Add "tags" array to JSON Feed 1.1 items
- Omit empty tags field in JSON Feed (minimal approach)

Standards compliance:
- RSS 2.0: category element with display name
- Atom 1.0: RFC 4287 category with term and label attributes
- JSON Feed 1.1: tags array of strings

Per design: docs/design/v1.3.1/feed-tags-design.md
Implementation: docs/design/v1.3.1/2025-12-10-feed-tags-implementation.md

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 12:33:35 -07:00
philandClaude 3222620cee release: Bump version to 1.3.0
Build Container / build (push) Failing after 12s
Promoting v1.3.0-rc.1 to stable release.

Changes:
- Updated version in starpunk/__init__.py to 1.3.0
- Updated CHANGELOG.md header to v1.3.0

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 12:19:11 -07:00
philandClaude 247eb34c36 Merge feature/v1.3.0-tags-microformats into main
Release candidate v1.3.0-rc.1 with tags/categories and enhanced Microformats2 support.

Major features:
- Complete tag/category system with Micropub support
- Strict Microformats2 compliance (p-category, h-feed properties)
- Tag archive pages at /tags/{tag}
- Enhanced h-entry markup with dt-updated
- Proper h-feed structure on collection pages

Generated with Claude Code

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 12:01:09 -07:00
philandClaude 41b65703f9 docs: Add v1.3.1 and v1.4.0 release definitions
v1.3.1 "Syndicate Tags":
- RSS/Atom/JSON Feed category/tag support

v1.4.0 "Media":
- Micropub media endpoint (W3C compliant)
- Large image support (>10MB auto-resize)
- Enhanced feed media (image variants, full Media RSS)

Also adds tag-filtered feeds to backlog at medium priority.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 12:00:56 -07:00
phil f901aa2242 docs: Update project plan files 2025-12-10 11:58:45 -07:00
philandClaude 5ca8b7e9b4 release: Bump version to 1.3.0-rc.1
Release candidate for v1.3.0 with tags/categories and enhanced Microformats2 support.

Features:
- Tag/category system with Micropub support
- Strict Microformats2 compliance (p-category, h-feed)
- Tag archive pages
- Enhanced h-entry markup

Generated with Claude Code

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 11:58:10 -07:00
philandClaude 3d80e1af51 test(microformats): Add v1.3.0 validation tests for tags and h-feed
Phase 4: Validation per microformats-tags-design.md

Added test fixtures:
- published_note_with_tags: Creates note with test tags for p-category validation
- published_note_with_media: Creates note with media for u-photo placement testing

Added v1.3.0 microformats2 validation tests:
- test_hfeed_has_required_properties: Validates name, author, url per spec
- test_hfeed_author_is_valid_hcard: Validates h-card structure
- test_hentry_has_pcategory_for_tags: Validates p-category markup
- test_uphoto_outside_econtent: Validates u-photo placement per draft spec

Test results:
- All 18 microformats tests pass
- All 116 related tests pass (microformats, notes, micropub)
- Confirms Phases 1-3 implementation correctness

Updated BACKLOG.md with tag-filtered feeds feature (medium priority)

Implementation report: docs/design/v1.3.0/2025-12-10-phase4-implementation.md

Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 11:51:39 -07:00
philandClaude 372064b116 feat(tags): Add tag archive route and admin interface integration
Implement Phase 3 of v1.3.0 tags feature per microformats-tags-design.md:

Routes (starpunk/routes/public.py):
- Add /tag/<tag> archive route with normalization and 404 handling
- Pre-load tags in index route for all notes
- Pre-load tags in note route for individual notes

Admin (starpunk/routes/admin.py):
- Parse comma-separated tag input in create route
- Parse tag input in update route
- Pre-load tags when displaying edit form
- Empty tag field removes all tags

Templates:
- Add tag input field to templates/admin/edit.html
- Add tag input field to templates/admin/new.html
- Use Jinja2 map filter to display existing tags

Implementation details:
- Tag URL parameter normalized to lowercase before lookup
- Tags pre-loaded using object.__setattr__ pattern (like media)
- parse_tag_input() handles trim, dedupe, normalization
- All existing tests pass (micropub categories, admin routes)

Per architect design:
- No pagination on tag archives (acceptable for v1.3.0)
- No autocomplete in admin (out of scope)
- Follows existing media loading patterns

Generated with Claude Code

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 11:42:16 -07:00
philandClaude 377027e79a feat(templates): Add microformats2 h-feed and p-category markup for tags
Implement Phase 2 of v1.3.0 per microformats-tags-design.md

Template Updates:
- templates/index.html: Add h-feed properties (u-url, enhanced p-author with u-photo/p-note, feed-level u-photo)
- templates/index.html: Add p-category markup with rel="tag" to note previews
- templates/note.html: Add p-category markup with rel="tag" for tags
- templates/note.html: Enhance author h-card with u-photo and p-note (hidden for parsers)
- templates/note.html: Document u-photo placement outside e-content per draft spec
- templates/tag.html: Create new tag archive template with h-feed structure

Key Decisions Applied:
- Tags ordered alphabetically by display_name (ready for backend)
- rel="tag" on all p-category links per microformats2 spec
- Author bio (p-note) hidden with display: none for semantic parsing
- Dual u-photo elements intentional for parser compatibility
- Graceful fallback when author photo/bio not available

Templates are backward compatible and ready for backend integration.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 11:35:11 -07:00
philandClaude f10d0679da feat(tags): Add database schema and tags module (v1.3.0 Phase 1)
Implements tag/category system backend following microformats2 p-category specification.

Database changes:
- Migration 008: Add tags and note_tags tables
- Normalized tag storage (case-insensitive lookup, display name preserved)
- Indexes for performance

New module:
- starpunk/tags.py: Tag management functions
  - normalize_tag: Normalize tag strings
  - get_or_create_tag: Get or create tag records
  - add_tags_to_note: Associate tags with notes (replaces existing)
  - get_note_tags: Retrieve note tags (alphabetically ordered)
  - get_tag_by_name: Lookup tag by normalized name
  - get_notes_by_tag: Get all notes with specific tag
  - parse_tag_input: Parse comma-separated tag input

Model updates:
- Note.tags property (lazy-loaded, prefer pre-loading in routes)
- Note.to_dict() add include_tags parameter

CRUD updates:
- create_note() accepts tags parameter
- update_note() accepts tags parameter (None = no change, [] = remove all)

Micropub integration:
- Pass tags to create_note() (tags already extracted by extract_tags())
- Return tags in q=source response

Per design doc: docs/design/v1.3.0/microformats-tags-design.md

Generated with Claude Code

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 11:24:23 -07:00
philandClaude 927db4aea0 release: Bump version to 1.2.0
Build Container / build (push) Failing after 1m52s
Promote v1.2.0-rc.2 to stable v1.2.0 release

- Merged rc.1 and rc.2 changelog entries
- Updated version in starpunk/__init__.py
- All features tested in production

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-10 08:39:54 -07:00
philandClaude 27501f6381 feat: v1.2.0-rc.2 - Media display fixes and feed enhancements
## Added
- Feed Media Enhancement with Media RSS namespace support
  - RSS enclosure, media:content, media:thumbnail elements
  - JSON Feed image field for first image
- ADR-059: Full feed media standardization roadmap

## Fixed
- Media display on homepage (was only showing on note pages)
- Responsive image sizing with CSS constraints
- Caption display (now alt text only, not visible)
- Logging correlation ID crash in non-request contexts

## Documentation
- Feed media design documents and implementation reports
- Media display fixes design and validation reports
- Updated ROADMAP with v1.3.0/v1.4.0 media plans

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-12-09 14:58:37 -07:00
phil 10d85bb78b fix: Apply correlation filter to handlers for proper multi-logger support
Fixes logging errors during app initialization and in background threads.
The correlation_id filter must be applied to handlers (not just loggers)
to ensure all log records have the correlation_id attribute before
formatting occurs.

Issue: Gunicorn workers were crashing due to missing correlation_id
in logs from memory monitor and other non-request contexts.
2025-11-28 16:22:12 -07:00
philandClaude dd822a35b5 feat: v1.2.0-rc.1 - IndieWeb Features Release Candidate
Complete implementation of v1.2.0 "IndieWeb Features" release.

## Phase 1: Custom Slugs
- Optional custom slug field in note creation form
- Auto-sanitization (lowercase, hyphens only)
- Uniqueness validation with auto-numbering
- Read-only after creation to preserve permalinks
- Matches Micropub mp-slug behavior

## Phase 2: Author Discovery + Microformats2
- Automatic h-card discovery from IndieAuth identity URL
- 24-hour caching with graceful fallback
- Never blocks login (per ADR-061)
- Complete h-entry, h-card, h-feed markup
- All required Microformats2 properties
- rel-me links for identity verification
- Passes IndieWeb validation

## Phase 3: Media Upload
- Upload up to 4 images per note (JPEG, PNG, GIF, WebP)
- Automatic optimization with Pillow
  - Auto-resize to 2048px
  - EXIF orientation correction
  - 95% quality compression
- Social media-style layout (media top, text below)
- Optional captions for accessibility
- Integration with all feed formats (RSS, ATOM, JSON Feed)
- Date-organized storage with UUID filenames
- Immutable caching (1 year)

## Database Changes
- migrations/006_add_author_profile.sql - Author discovery cache
- migrations/007_add_media_support.sql - Media storage

## New Modules
- starpunk/author_discovery.py - h-card discovery and caching
- starpunk/media.py - Image upload, validation, optimization

## Documentation
- 4 new ADRs (056, 057, 058, 061)
- Complete design specifications
- Developer Q&A with 40+ questions answered
- 3 implementation reports
- 3 architect reviews (all approved)

## Testing
- 56 new tests for v1.2.0 features
- 842 total tests in suite
- All v1.2.0 feature tests passing

## Dependencies
- Added: mf2py (Microformats2 parser)
- Added: Pillow (image processing)

Version: 1.2.0-rc.1

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-28 15:02:20 -07:00
philandClaude 83739ec2c6 release: Promote v1.1.2-rc.2 to stable v1.1.2 "Syndicate"
Build Container / build (push) Failing after 1m54s
Promoting release candidate to stable production release.

v1.1.2 "Syndicate" - Enhanced Content Distribution

This release delivers comprehensive metrics instrumentation and multi-format
feed support (RSS, ATOM, JSON Feed) with content negotiation, caching, and
statistics dashboard.

No changes from v1.1.2-rc.2 - both production issues verified fixed.

Version: 1.1.2 (stable)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-28 09:59:42 -07:00
philandClaude 1e2135a49a fix: Resolve v1.1.2-rc.1 production issues - Static files and metrics
This release candidate fixes two critical production issues discovered in v1.1.2-rc.1:

1. CRITICAL: Static files returning 500 errors
   - HTTP monitoring middleware was accessing response.data on streaming responses
   - Fixed by checking direct_passthrough flag before accessing response data
   - Static files (CSS, JS, images) now load correctly
   - File: starpunk/monitoring/http.py

2. HIGH: Database metrics showing zero
   - Configuration key mismatch: config set METRICS_SAMPLING_RATE (singular),
     buffer read METRICS_SAMPLING_RATES (plural)
   - Fixed by standardizing on singular key name
   - Modified MetricsBuffer to accept both float and dict for flexibility
   - Changed default sampling from 10% to 100% for better visibility
   - Files: starpunk/monitoring/metrics.py, starpunk/config.py

Version: 1.1.2-rc.2

Documentation:
- Investigation report: docs/reports/2025-11-28-v1.1.2-rc.1-production-issues.md
- Architect review: docs/reviews/2025-11-28-v1.1.2-rc.1-architect-review.md
- Implementation report: docs/reports/2025-11-28-v1.1.2-rc.2-fixes.md

Testing: All monitoring tests pass (28/28)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-28 09:46:31 -07:00
philandClaude 34b576ff79 docs: Add upgrade guide for v1.1.2-rc.1
🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-28 02:12:24 -07:00
philandClaude dd63df7858 chore: Bump version to 1.1.2-rc.1
🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-28 02:03:46 -07:00
philandClaude 7dc2f11670 Merge v1.1.2 Phase 3 - Feed Enhancements (Caching, Statistics, OPML)
Completes the v1.1.2 "Syndicate" release with feed enhancements.

Phase 3 Deliverables:
- Feed caching with LRU + TTL (5 minutes)
- ETag support with 304 Not Modified responses
- Feed statistics dashboard integration
- OPML 2.0 export endpoint

Features:
- LRU cache with SHA-256 checksums
- Weak ETags for bandwidth optimization
- Feed format statistics and cache efficiency metrics
- OPML subscription list at /opml.xml
- Feed discovery link in HTML

Quality Metrics:
- 766 total tests passing (100%)
- Zero breaking changes
- Cache bounded at 50 entries
- <1ms caching overhead
- Production-ready

Architect Review: APPROVED WITH COMMENDATIONS (10/10)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-27 21:44:44 -07:00
philandClaude 32fe1de50f feat: Complete v1.1.2 Phase 3 - Feed Enhancements (Caching, Statistics, OPML)
Implements caching, statistics, and OPML export for multi-format feeds.

Phase 3 Deliverables:
- Feed caching with LRU + TTL (5 minutes)
- ETag support with 304 Not Modified responses
- Feed statistics dashboard integration
- OPML 2.0 export endpoint

Features:
- LRU cache with SHA-256 checksums for weak ETags
- 304 Not Modified responses for bandwidth optimization
- Feed format statistics tracking (RSS, ATOM, JSON Feed)
- Cache efficiency metrics (hit/miss rates, memory usage)
- OPML subscription list at /opml.xml
- Feed discovery link in HTML base template

Quality Metrics:
- All existing tests passing (100%)
- Cache bounded at 50 entries with 5-minute TTL
- <1ms caching overhead
- Production-ready implementation

Architect Review: APPROVED WITH COMMENDATIONS (10/10)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-27 21:42:37 -07:00
philandClaude c1dd706b8f feat: Implement Phase 3 Feed Caching (Partial)
Implements feed caching layer with LRU eviction, TTL expiration, and ETag support.

Phase 3.1: Feed Caching (Complete)
- LRU cache with configurable max_size (default: 50 feeds)
- TTL-based expiration (default: 300 seconds = 5 minutes)
- SHA-256 checksums for cache keys and ETags
- Weak ETag generation (W/"checksum")
- If-None-Match header support for 304 Not Modified responses
- Cache invalidation (全体 or per-format)
- Hit/miss/eviction statistics tracking
- Content-based cache keys (changes when notes are modified)

Implementation:
- Created starpunk/feeds/cache.py with FeedCache class
- Integrated caching into feed routes (RSS, ATOM, JSON Feed)
- Added ETag headers to all feed responses
- 304 Not Modified responses for conditional requests
- Configuration: FEED_CACHE_ENABLED, FEED_CACHE_MAX_SIZE
- Global cache instance with singleton pattern

Architecture:
- Two-level caching:
  1. Note list cache (simple dict, existing)
  2. Feed content cache (LRU with TTL, new)
- Cache keys include format + notes checksum
- Checksums based on note IDs + updated timestamps
- Non-streaming generators used for cacheable content

Testing:
- 25 comprehensive cache tests (100% passing)
- Tests for LRU eviction, TTL expiration, statistics
- Tests for checksum generation and consistency
- Tests for ETag generation and uniqueness
- All 114 feed tests passing (no regressions)

Quality Metrics:
- 114/114 tests passing (100%)
- Zero breaking changes
- Full backward compatibility
- Cache disabled mode supported (FEED_CACHE_ENABLED=false)

Performance Benefits:
- Database queries reduced (note list cached)
- Feed generation reduced (content cached)
- Bandwidth saved (304 responses)
- Memory efficient (LRU eviction)

Note: Phase 3 is partially complete. Still pending:
- Feed statistics dashboard
- OPML 2.0 export endpoint

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-27 21:14:03 -07:00
philandClaude f59cbb30a5 Merge v1.1.2 Phase 2 - Feed Formats (RSS, ATOM, JSON Feed)
Implements multiple feed format support with content negotiation.

Phase 2 Deliverables:
- Phase 2.0: Fixed RSS ordering regression (oldest-first → newest-first)
- Phase 2.1: Restructured feeds into modular package
- Phase 2.2: ATOM 1.0 feed implementation (RFC 4287)
- Phase 2.3: JSON Feed 1.1 implementation
- Phase 2.4: HTTP content negotiation with 5 endpoints

Feed Formats:
- RSS 2.0: Fully compliant, streaming + non-streaming
- ATOM 1.0: RFC 4287 compliant, RFC 3339 dates
- JSON Feed 1.1: Spec compliant with custom extension

Endpoints:
- /feed - Content negotiation via Accept header
- /feed.rss - Explicit RSS 2.0
- /feed.atom - Explicit ATOM 1.0
- /feed.json - Explicit JSON Feed 1.1
- /feed.xml - Backward compatibility (→ RSS)

Quality Metrics:
- 111/111 feed tests passing (100%)
- Zero breaking changes
- Full backward compatibility
- Standards compliant (RSS 2.0, ATOM 1.0, JSON Feed 1.1)
- Performance: 2-5ms generation per 50 items

Architect Review: APPROVED WITH COMMENDATION

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-27 20:58:33 -07:00
philandClaude 8fbdcb6e6f feat: Complete Phase 2.4 - HTTP Content Negotiation
Implements HTTP content negotiation for feed format selection.

Phase 2.4 Deliverables:
- Content negotiation via Accept header parsing
- Quality factor support (q= parameter)
- 5 feed endpoints with format routing
- 406 Not Acceptable responses with helpful errors
- Comprehensive test coverage (63 tests)

Endpoints:
- /feed - Content negotiation based on Accept header
- /feed.rss - Explicit RSS 2.0
- /feed.atom - Explicit ATOM 1.0
- /feed.json - Explicit JSON Feed 1.1
- /feed.xml - Backward compatibility (→ RSS)

MIME Type Mapping:
- application/rss+xml → RSS 2.0
- application/atom+xml → ATOM 1.0
- application/feed+json or application/json → JSON Feed 1.1
- */* → RSS 2.0 (default)

Implementation:
- Simple quality factor parsing (StarPunk philosophy)
- Not full RFC 7231 compliance (minimal approach)
- Reuses existing feed generators
- No breaking changes

Quality Metrics:
- 132/132 tests passing (100%)
- Zero breaking changes
- Full backward compatibility
- Standards compliant negotiation

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-27 20:46:49 -07:00
philandClaude 59e9d402c6 feat: Implement Phase 2 Feed Formats - ATOM, JSON Feed, RSS fix (Phases 2.0-2.3)
This commit implements the first three phases of v1.1.2 Phase 2 Feed Formats,
adding ATOM 1.0 and JSON Feed 1.1 support alongside the existing RSS feed.

CRITICAL BUG FIX:
- Fixed RSS streaming feed ordering (was showing oldest-first instead of newest-first)
- Streaming RSS removed incorrect reversed() call at line 198
- Feedgen RSS kept correct reversed() to compensate for library behavior

NEW FEATURES:
- ATOM 1.0 feed generation (RFC 4287 compliant)
  - Proper XML namespacing and RFC 3339 dates
  - Streaming and non-streaming methods
  - 11 comprehensive tests

- JSON Feed 1.1 generation (JSON Feed spec compliant)
  - RFC 3339 dates and UTF-8 JSON output
  - Custom _starpunk extension with permalink_path and word_count
  - 13 comprehensive tests

REFACTORING:
- Restructured feed code into starpunk/feeds/ module
  - feeds/rss.py - RSS 2.0 (moved from feed.py)
  - feeds/atom.py - ATOM 1.0 (new)
  - feeds/json_feed.py - JSON Feed 1.1 (new)
- Backward compatible feed.py shim for existing imports
- Business metrics integrated into all feed generators

TESTING:
- Created shared test helper tests/helpers/feed_ordering.py
- Helper validates newest-first ordering across all formats
- 48 total feed tests, all passing
  - RSS: 24 tests
  - ATOM: 11 tests
  - JSON Feed: 13 tests

FILES CHANGED:
- Modified: starpunk/feed.py (now compatibility shim)
- New: starpunk/feeds/ module with rss.py, atom.py, json_feed.py
- New: tests/helpers/feed_ordering.py (shared test helper)
- New: tests/test_feeds_atom.py, tests/test_feeds_json.py
- Modified: CHANGELOG.md (Phase 2 entries)
- New: docs/reports/2025-11-26-v1.1.2-phase2-feed-formats-partial.md

NEXT STEPS:
Phase 2.4 (Content Negotiation) pending - will add /feed endpoint with
Accept header negotiation and explicit format endpoints.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-26 14:54:52 -07:00
philandClaude a99b27d4e9 Merge v1.1.2 Phase 1 - Complete Metrics Instrumentation
Implements the metrics instrumentation that was missing from v1.1.1.
The monitoring framework existed but was never actually used to collect metrics.

Phase 1 Deliverables:
- Database operation monitoring with query timing
- HTTP request/response metrics with request IDs
- Memory monitoring daemon thread
- Business metrics framework
- Configuration management

Quality Metrics:
- 28/28 tests passing (100%)
- Zero architectural deviations
- <1% performance overhead achieved
- Production-ready implementation

Architect Review: APPROVED with excellent marks

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-26 14:14:54 -07:00
philandClaude b0230b1233 feat: Complete v1.1.2 Phase 1 - Metrics Instrumentation
Implements the metrics instrumentation framework that was missing from v1.1.1.
The monitoring framework existed but was never actually used to collect metrics.

Phase 1 Deliverables:
- Database operation monitoring with query timing and slow query detection
- HTTP request/response metrics with request IDs for all requests
- Memory monitoring via daemon thread with configurable intervals
- Business metrics framework for notes, feeds, and cache operations
- Configuration management with environment variable support

Implementation Details:
- MonitoredConnection wrapper at pool level for transparent DB monitoring
- Flask middleware hooks for HTTP metrics collection
- Background daemon thread for memory statistics (skipped in test mode)
- Simple business metric helpers for integration in Phase 2
- Comprehensive test suite with 28/28 tests passing

Quality Metrics:
- 100% test pass rate (28/28 tests)
- Zero architectural deviations from specifications
- <1% performance overhead achieved
- Production-ready with minimal memory impact (~2MB)

Architect Review: APPROVED with excellent marks

Documentation:
- Implementation report: docs/reports/v1.1.2-phase1-metrics-implementation.md
- Architect review: docs/reviews/2025-11-26-v1.1.2-phase1-review.md
- Updated CHANGELOG.md with Phase 1 additions

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
2025-11-26 14:13:44 -07:00
593 changed files with 64607 additions and 17052 deletions

No files matched your search

+208 -70
View File
@@ -1,102 +1,240 @@
# StarPunk Configuration Template
# Copy this file to .env and fill in your values
# DO NOT commit .env to version control
# StarPunk configuration template (v2.14.0)
#
# Copy to .env and fill in your values. DO NOT commit .env.
# Every variable here is read by starpunk/config.py. Uncommented lines are
# the ones you normally set; commented lines show the built-in default and
# only need uncommenting to change it. Real environment variables take
# precedence over .env.
# =============================================================================
# SITE CONFIGURATION
# SITE
# =============================================================================
# Public URL where your site is hosted (no trailing slash)
# Public URL of the site (a trailing slash is added if missing)
# (default: http://localhost:5000)
SITE_URL=http://localhost:5000
# Your site name (appears in RSS feed and page titles)
# Site name: page titles and feeds (default: StarPunk)
SITE_NAME=My StarPunk Site
# Your name (appears as author in RSS feed)
# Author name; must be non-empty but is not currently displayed: pages use
# the h-card discovered at ADMIN_ME (default: Unknown)
SITE_AUTHOR=Your Name
# Site description (appears in RSS feed)
# Feed description (default: A minimal IndieWeb CMS)
SITE_DESCRIPTION=My personal IndieWeb site
# =============================================================================
# AUTHENTICATION
# =============================================================================
# Your IndieWeb identity URL (REQUIRED)
# This is YOUR personal website URL that you authenticate with
# Example: https://yourname.com or https://github.com/yourname
# Your identity URL (REQUIRED unless DEV_MODE=true). Used as the author h-card
# source and as the "me" of Micropub tokens. It should link to this site's
# IndieAuth metadata so Micropub apps can find StarPunk (ADR-064).
ADMIN_ME=https://your-website.com
# Session secret key (REQUIRED - GENERATE A RANDOM VALUE)
# Session secret (REQUIRED: the app refuses to start without it)
# Generate with: python3 -c "import secrets; print(secrets.token_hex(32))"
SESSION_SECRET=REPLACE_WITH_RANDOM_SECRET
# Session lifetime in days (default: 30)
SESSION_LIFETIME=30
# Flask secret key (default: SESSION_SECRET)
# FLASK_SECRET_KEY=
# IndieLogin service URL (usually don't change this)
INDIELOGIN_URL=https://indielogin.com
# Admin session lifetime in days (default: 30)
# SESSION_LIFETIME=30
# Admin login via OpenID Connect (all four REQUIRED unless DEV_MODE=true)
# Register StarPunk as a confidential client with your provider (e.g. Pocket ID)
# using the callback URL: <SITE_URL>/auth/callback
OIDC_ISSUER=https://id.example.com
OIDC_CLIENT_ID=
OIDC_CLIENT_SECRET=
# The provider's subject ("sub") of the one account allowed to log in.
# Pocket ID: the user's ID shown in the admin user list.
OIDC_ADMIN_SUB=
# Name shown on the login button (default: single sign-on)
OIDC_PROVIDER_NAME=Pocket ID
# Lifetime of the Micropub access tokens StarPunk's IndieAuth server issues,
# in days (default: 90)
# INDIEAUTH_TOKEN_LIFETIME_DAYS=90
# =============================================================================
# DATA STORAGE
# DATA STORAGE (the container image sets /data, /data/notes, /data/starpunk.db)
# =============================================================================
# Base data directory (relative to project root)
# Base data directory: notes (DATA_PATH/notes), media, logs,
# threads_token.json (default: ./data)
DATA_PATH=./data
# Notes directory (where markdown files are stored)
# Created at startup; note files are always written under DATA_PATH/notes,
# so keep this equal to that (default: ./data/notes)
NOTES_PATH=./data/notes
# SQLite database path
# SQLite database (default: ./data/starpunk.db)
DATABASE_PATH=./data/starpunk.db
# =============================================================================
# FLASK CONFIGURATION
# FLASK / LOGGING
# =============================================================================
# Environment: development or production
FLASK_ENV=development
# Debug mode: 1 (on) or 0 (off)
# NEVER use debug mode in production
FLASK_DEBUG=1
# Flask secret key (falls back to SESSION_SECRET if not set)
FLASK_SECRET_KEY=
# =============================================================================
# RSS FEED CONFIGURATION
# =============================================================================
# Maximum number of items in RSS feed (default: 50)
FEED_MAX_ITEMS=50
# Feed cache duration in seconds (default: 300 = 5 minutes)
FEED_CACHE_SECONDS=300
# =============================================================================
# CONTAINER CONFIGURATION
# =============================================================================
# Environment: development or production
ENVIRONMENT=production
# Number of Gunicorn workers (default: 4)
# Recommendation: (2 x CPU cores) + 1
WORKERS=4
# Worker timeout in seconds (default: 30)
WORKER_TIMEOUT=30
# Max requests per worker before restart (prevents memory leaks)
MAX_REQUESTS=1000
# =============================================================================
# DEVELOPMENT OPTIONS
# =============================================================================
# Log level: DEBUG, INFO, WARNING, ERROR, CRITICAL
# development or production (default: production)
FLASK_ENV=production
# 1 = debug on, 0 = off (default: 0). `flask run --debug` turns it on locally.
FLASK_DEBUG=0
# DEBUG, INFO, WARNING, ERROR or CRITICAL (default: INFO)
LOG_LEVEL=INFO
# Overrides the version shown in the footer and /health (default: the
# package version, starpunk.__version__). Normally leave unset.
# VERSION=
# Enable SQL query logging (development only)
SQL_ECHO=0
# =============================================================================
# DEVELOPMENT LOGIN (never in production)
# =============================================================================
# DEV_MODE=true adds /dev/login, which logs you in as DEV_ADMIN_ME without
# OIDC, and makes ADMIN_ME and OIDC_* optional (default: false)
# DEV_MODE=false
# Identity used by /dev/login (required when DEV_MODE=true)
# DEV_ADMIN_ME=http://localhost:5000/
# =============================================================================
# REVERSE PROXY
# =============================================================================
# Behind a reverse proxy (Caddy, nginx): take the client address from
# X-Forwarded-For. Without it every visitor shares the proxy's address, so one
# busy client can rate-limit everyone on /webmention and /websub.
# Only enable when the proxy sets X-Forwarded-For itself (Caddy does).
# (default: false)
TRUSTED_PROXY=false
# CSRF tokens on admin forms, logout and the IndieAuth consent (default: true).
# Leave on; tests turn it off unless they test it.
CSRF_ENABLED=true
# =============================================================================
# FEEDS
# =============================================================================
# Items per feed (default: 50)
# FEED_MAX_ITEMS=50
# Feed cache lifetime in seconds; changes to notes clear it (default: 300)
# FEED_CACHE_SECONDS=300
# FEED_CACHE_ENABLED=true
# Cached feed variants kept (default: 50)
# FEED_CACHE_MAX_SIZE=50
# =============================================================================
# WEBMENTION
# =============================================================================
# Webmention endpoint at /webmention, its Link header, and showing responses
# (webmentions and backfeed) on notes; sending is not affected (default: true)
# WEBMENTION_ENABLED=true
# Background worker that verifies received webmentions, sends outgoing ones,
# and sends response notifications (v2.14.0) (default: true)
# WEBMENTION_WORKER_ENABLED=true
# Hosts never sent webmentions, comma-separated, e.g. self-hosted apps that
# don't accept them. Links to this site itself are always skipped (v2.18.3)
# WEBMENTION_SKIP_HOSTS=git.example.com,photos.example.com
# Seconds between queue checks (default: 10) and after an error (default: 60)
# WEBMENTION_POLL_INTERVAL=10
# WEBMENTION_ERROR_BACKOFF=60
# Items per worker pass (default: 5)
# WEBMENTION_BATCH_SIZE=5
# Show verified mentions without moderation (default: true)
# WEBMENTION_AUTO_APPROVE=true
# Rate limits on /webmention: requests per window per client IP (default: 60)
# and per source domain (default: 10); window in seconds (default: 60)
# WEBMENTION_RATE_LIMIT_ENABLED=true
# WEBMENTION_RATE_LIMIT_IP=60
# WEBMENTION_RATE_LIMIT_DOMAIN=10
# WEBMENTION_RATE_LIMIT_WINDOW=60
# =============================================================================
# WEBSUB (v2.10.0)
# =============================================================================
# Built-in hub at /websub for this site's feeds. Set WEBSUB_HUB_URL to use
# another hub instead (StarPunk then pings it on changes). (default: true)
WEBSUB_ENABLED=true
# WEBSUB_HUB_URL=https://hub.example.com/
# Subscriptions in total, pending included (default: 1000)
# WEBSUB_MAX_SUBSCRIPTIONS=1000
# Per callback server (resolved address) (default: 25)
# WEBSUB_MAX_PER_HOST=25
# Seconds between hub worker passes (default: 5)
# WEBSUB_POLL_INTERVAL=5
# =============================================================================
# CROSSPOSTING / POSSE (optional; a target is enabled when its credentials
# are set)
# =============================================================================
# GoToSocial or any Mastodon-API server (v2.1.0, ADR-065). Create an app
# token with scopes: write:statuses write:media, plus read:search
# write:favourites for native likes and boosts (v2.5.0), plus read:statuses
# read:accounts (or read) for backfeed (v2.13.0)
GOTOSOCIAL_URL=
GOTOSOCIAL_ACCESS_TOKEN=
# Name shown in the editor and Micropub apps (default: GoToSocial)
# GOTOSOCIAL_NAME=GoToSocial
# public, unlisted or private (default: public)
# GOTOSOCIAL_VISIBILITY=public
# Server limits (defaults: 5000 characters, 4 images)
# GOTOSOCIAL_MAX_CHARS=5000
# GOTOSOCIAL_MAX_MEDIA=4
# Bluesky (v2.2.0, ADR-066). Use an app password, not your account password.
BLUESKY_HANDLE=
BLUESKY_APP_PASSWORD=
# PDS / entryway (default: https://bsky.social)
# BLUESKY_SERVICE=https://bsky.social
# BLUESKY_NAME=Bluesky
# BLUESKY_MAX_CHARS=300
# Threads (v2.6.0, ADR-069). A long-lived user token with threads_basic,
# threads_content_publish and threads_delete (plus threads_read_replies and
# threads_manage_insights for backfeed, v2.13.0); StarPunk refreshes it daily and
# keeps the current one in DATA_PATH/threads_token.json.
# How to get one: docs/design/v2.6.0/2026-10-31-threads-crossposting.md
THREADS_ACCESS_TOKEN=
# Your Threads username (for the Micropub syndicate-to uid)
THREADS_USERNAME=
# THREADS_NAME=Threads
# THREADS_MAX_CHARS=500
# THREADS_MAX_MEDIA=4
# Seconds between syndication worker passes (default: 10)
# SYNDICATION_POLL_INTERVAL=10
# Backfeed (v2.13.0, ADR-073): bring likes, reposts, quotes and replies to the
# crossposts back to the note as responses. Needs the read scopes above, and
# runs in the syndication worker, so at least one target must be configured.
# BACKFEED_ENABLED=false
# Stop checking a note's responses after this many days (default: 90)
# BACKFEED_MAX_AGE_DAYS=90
# =============================================================================
# NOTIFICATIONS (v2.14.0, ADR-074; optional)
# =============================================================================
# Apprise URLs to notify, space- or comma-separated
# (https://github.com/caronc/apprise/wiki). Empty: notifications off.
# Examples: gotifys://gotify.example.com/APPTOKEN ntfys://my-topic
# apprises://apprise.example.com/KEY (an apprise-api server)
# NOTIFY_URLS=
# What to notify: reply, mention, quote, like, repost, bookmark, problems
# (a crosspost that failed for good, a silo refusing backfeed). Names are
# case-insensitive; unknown ones are ignored.
# NOTIFY_EVENTS=reply,mention,quote,problems
# Priority of every message, added to URLs that don't set one:
# low, moderate, normal, high, emergency (Gotify also takes 0-10)
# NOTIFY_PRIORITY=normal
# =============================================================================
# DEBUGGING
# =============================================================================
# Keep uploads that fail processing in DATA_PATH/debug for diagnosis
# (default: false); cleaned up by age (default: 7 days) and total size
# (default: 100 MB)
# DEBUG_SAVE_FAILED_UPLOADS=false
# DEBUG_FILE_MAX_AGE_DAYS=7
# DEBUG_FILE_MAX_SIZE_MB=100
+30 -42
View File
@@ -1,58 +1,46 @@
# Gitea Actions workflow for StarPunk
# Builds and pushes container images on version tags
# Runs the tests, then builds and pushes container images on version tags.
#
# The runner has no Docker daemon (that would give workflows root on the
# server). Images are built by the rootless BuildKit container next to it,
# reached through its socket.
name: Build Container
on:
# Trigger on version tags
push:
tags:
- 'v[0-9]+.[0-9]+.[0-9]+'
# Allow manual trigger from Gitea UI
workflow_dispatch:
env:
REGISTRY: git.thesatelliteoflove.com
BUILDKIT: unix:///run/buildkit/buildkit/buildkitd.sock
jobs:
build:
runs-on: docker
steps:
- name: Install dependencies
run: |
if command -v apk > /dev/null; then
apk add --no-cache nodejs npm docker git
elif command -v apt-get > /dev/null; then
apt-get update && apt-get install -y nodejs npm docker.io git
elif command -v yum > /dev/null; then
yum install -y nodejs npm docker git
fi
run: apk add --no-cache nodejs git docker-cli docker-cli-buildx
- name: Checkout
uses: actions/checkout@v4
- name: Set up Docker Buildx
uses: docker/setup-buildx-action@v3
- name: Extract registry URL
id: registry
run: |
# Extract hostname from server URL (remove protocol)
REGISTRY_URL=$(echo "${{ github.server_url }}" | sed 's|https://||' | sed 's|http://||')
echo "url=${REGISTRY_URL}" >> $GITHUB_OUTPUT
- name: Login to Gitea Container Registry
uses: docker/login-action@v3
with:
registry: ${{ steps.registry.outputs.url }}
username: ${{ github.actor }}
password: ${{ secrets.GITHUB_TOKEN }}
# The automatic Actions token can't log in to the container registry, so
# this uses a personal access token with only the write:package scope,
# stored as the repository secret REGISTRY_TOKEN.
- name: Log in to the Gitea registry
run: echo "${{ secrets.REGISTRY_TOKEN }}" | docker login "$REGISTRY" -u "${{ github.repository_owner }}" --password-stdin
- name: Use the BuildKit socket
run: docker buildx create --name ci --driver remote "$BUILDKIT" --use
# Builds the Containerfile's test stage, which runs pytest; a failure
# stops the workflow before anything is pushed
- name: Run tests
run: docker buildx build --target test --progress plain -f Containerfile .
- name: Build and push
uses: docker/build-push-action@v5
with:
context: .
file: ./Containerfile
push: true
tags: |
${{ steps.registry.outputs.url }}/${{ github.repository }}:${{ github.ref_name }}
${{ steps.registry.outputs.url }}/${{ github.repository }}:latest
cache-from: type=gha
cache-to: type=gha,mode=max
run: |
image="$REGISTRY/$(echo '${{ github.repository }}' | tr '[:upper:]' '[:lower:]')"
docker buildx build --push -f Containerfile \
-t "$image:${{ github.ref_name }}" -t "$image:latest" .
- name: Clean up
if: always()
run: |
docker buildx rm ci || true
docker logout "$REGISTRY" || true
+1297 -4
View File
File diff suppressed because it is too large. Load diff
+30 -67
View File
@@ -6,99 +6,62 @@ This file contains operational instructions for Claude agents working on this pr
- We use **uv** for Python virtual environment management
- All Python commands must be run with `uv run` prefix
- Example: `uv run pytest`, `uv run flask run`
- Example: `uv run pytest`, `uv run flask --app app.py run --debug`
- There is no `pyproject.toml`: create the environment with `uv venv --python 3.14 && uv pip install -r requirements-dev.txt`
- Tests need `SESSION_SECRET` set. Locally, run only the tests for the area you changed (`SESSION_SECRET=test uv run pytest tests/test_<area>.py`); CI runs the full suite (about 6 minutes) on every `vX.Y.Z` tag, so don't run the whole suite yourself unless asked
- Local login without OIDC: `DEV_MODE=true` and `DEV_ADMIN_ME` in `.env`, then `/dev/login`
- Setup details: `docs/standards/development-setup.md`
## Agent Protocol (All Agents)
**IMPORTANT**: All agents must review `docs/DOCUMENTATION.md` before starting work. This file is the authoritative source for documentation organization and supersedes any other instructions.
## Agent-Architect Protocol
When invoking the agent-architect, always remind it to:
1. Review documentation in docs/ before working on the task it is given
- docs/architecture, docs/decisions, docs/standards are of particular interest
1. Review `docs/DOCUMENTATION.md` for documentation organization standards
2. Give it the map of the documentation folder as described in the "Understanding the docs/ Structure" section below
2. Review documentation in docs/ before working on the task it is given
- docs/architecture, docs/decisions, docs/standards are of particular interest
3. Search for authoritative documentation for any web standard it is implementing on https://www.w3.org/
4. If it is reviewing a developers implementation report and it is accepts the completed work it should go back and update the project plan to reflect the completed work
4. If it is reviewing a developers implementation report and it accepts the completed work it should go back and update the project plan to reflect the completed work
## Agent-Developer Protocol
When invoking the agent-developer, always remind it to:
1. **Document work in reports**
- Create implementation reports in `docs/reports/`
- Include date in filename: `YYYY-MM-DD-description.md`
1. Review `docs/DOCUMENTATION.md` for documentation organization standards
2. **Update the changelog**
2. **Document work in design folder**
- Create implementation reports in `docs/design/{version}/`
- Include date in filename: `YYYY-MM-DD-description.md`
- All developer interaction (questions, responses, reports, reviews) goes in design/{version}/
3. **Update the changelog**
- Add entries to `CHANGELOG.md` for user-facing changes
- Follow existing format
3. **Version number management**
4. **Version number management**
- Increment version numbers according to `docs/standards/versioning-strategy.md`
- Update version in `starpunk/__init__.py`
- Update version in `starpunk/__init__.py` (both `__version__` and `__version_info__`)
4. **Follow git protocol**
5. **Follow git protocol**
- Adhere to git branching strategy in `docs/standards/git-branching-strategy.md`
- Create feature branches for non-trivial changes
- Write clear commit messages
- Create feature branches for non-trivial changes, in their own git worktree (`<prefix>/vX.Y.Z-description`, prefixes `feat/`, `fix/`, `ci/`, `chore/`, `docs/`)
- Write clear commit messages (`feat: ... (vX.Y.Z)`)
## Documentation Navigation
## Release Process
### Understanding the docs/ Structure
Follow "Release process" in `docs/standards/git-branching-strategy.md`. In short: work in a worktree on a `feat/`/`fix/`/`chore/` branch → bump `__version__` and `__version_info__` → `CHANGELOG.md` entry → design note in `docs/design/vX.Y.Z/` (or ADR amendment) → tests for the changed area → commit → `git merge --no-ff` into `main` ("Merge branch '<branch>' for vX.Y.Z release") → annotated tag `vX.Y.Z` → push `main` and the tag. CI (`.gitea/workflows/build-container.yml`) runs the full test suite and publishes the image; production is then updated from the separate Ansible deployment repository. Back up the database before deploying a release with a migration.
The `docs/` folder is organized by document type and purpose:
Merging to `main`, tagging, pushing and deploying happen only when the project owner asks for a release.
- **`docs/architecture/`** - System design overviews, component diagrams, architectural patterns
- **`docs/decisions/`** - Architecture Decision Records (ADRs), numbered sequentially (ADR-001, ADR-002, etc.)
- **`docs/deployment/`** - Deployment guides, infrastructure setup, operations documentation
- **`docs/design/`** - Detailed design documents, feature specifications, phase plans
- **`docs/examples/`** - Example implementations, code samples, usage patterns
- **`docs/migration/`** - Migration guides for upgrading between versions and configuration changes
- **`docs/projectplan/`** - Project roadmaps, implementation plans, feature scope definitions
- **`docs/releases/`** - Release-specific documentation, release notes, version information
- **`docs/reports/`** - Implementation reports from developers (dated: YYYY-MM-DD-description.md)
- **`docs/reviews/`** - Architectural reviews, design critiques, retrospectives
- **`docs/security/`** - Security-related documentation, vulnerability analyses, best practices
- **`docs/standards/`** - Coding standards, conventions, processes, workflows
## Documentation
### Where to Find Documentation
- **Before implementing a feature**: Check `docs/decisions/` for relevant ADRs and `docs/design/` for specifications
- **Understanding system architecture**: Start with `docs/architecture/overview.md`
- **Coding guidelines**: See `docs/standards/` for language-specific standards and best practices
- **Past implementation context**: Review `docs/reports/` for similar work (sorted by date)
- **Project roadmap and scope**: Refer to `docs/projectplan/`
### Where to Create New Documentation
**Create an ADR (`docs/decisions/`)** when:
- Making architectural decisions that affect system design
- Choosing between competing technical approaches
- Establishing patterns that others should follow
- Format: `ADR-NNN-brief-title.md` (find next number sequentially)
**Create a design doc (`docs/design/`)** when:
- Planning a complex feature implementation
- Detailing technical specifications
- Documenting multi-phase development plans
**Create an implementation report (`docs/reports/`)** when:
- Completing significant development work
- Documenting implementation details for architect review
- Format: `YYYY-MM-DD-brief-description.md`
**Update standards (`docs/standards/`)** when:
- Establishing new coding conventions
- Documenting processes or workflows
- Creating checklists or guidelines
### Key Documentation References
- **Architecture**: See `docs/architecture/overview.md`
- **Implementation Plan**: See `docs/projectplan/v1/implementation-plan.md`
- **Feature Scope**: See `docs/projectplan/v1/feature-scope.md`
- **Coding Standards**: See `docs/standards/python-coding-standards.md`
- **Testing**: See `docs/standards/testing-checklist.md`
See `docs/DOCUMENTATION.md` for the authoritative documentation structure, navigation guidance, and key references.
## Project Philosophy
+29 -4
View File
@@ -6,10 +6,10 @@
# ============================================================================
# Build Stage - Install dependencies in virtual environment
# ============================================================================
FROM python:3.11-slim AS builder
FROM python:3.14-slim AS builder
# Install uv for fast dependency installation
COPY --from=ghcr.io/astral-sh/uv:latest /uv /usr/local/bin/uv
COPY --from=ghcr.io/astral-sh/uv:0.9.28 /uv /usr/local/bin/uv
WORKDIR /build
@@ -23,9 +23,30 @@ RUN uv venv /opt/venv && \
uv pip install --no-cache -r requirements.txt
# ============================================================================
# Runtime Stage - Minimal production image
# Test Stage - Runs the test suite (CI builds it with --target test)
# ============================================================================
FROM python:3.11-slim
FROM python:3.14-slim AS test
COPY --from=ghcr.io/astral-sh/uv:0.9.28 /uv /usr/local/bin/uv
COPY --from=builder /opt/venv /opt/venv
COPY requirements-test.txt .
RUN uv pip install --no-cache --python /opt/venv/bin/python -r requirements-test.txt
ENV PATH="/opt/venv/bin:$PATH" \
PYTHONUNBUFFERED=1 \
PYTHONDONTWRITEBYTECODE=1 \
SESSION_SECRET=ci-test-secret \
SITE_URL=http://localhost:5000 \
ADMIN_ME=https://test.example.com
WORKDIR /app
COPY . .
RUN python -m pytest -q -p no:cacheprovider tests
# ============================================================================
# Runtime Stage - Minimal production image (default target: keep it last)
# ============================================================================
FROM python:3.14-slim
# Create non-root user for security
# UID/GID 1000 is standard for first user on most systems
@@ -40,6 +61,8 @@ COPY --from=builder /opt/venv /opt/venv
ENV PATH="/opt/venv/bin:$PATH" \
PYTHONUNBUFFERED=1 \
PYTHONDONTWRITEBYTECODE=1 \
FLASK_ENV=production \
FLASK_DEBUG=0 \
FLASK_APP=app.py \
DATA_PATH=/data \
NOTES_PATH=/data/notes \
@@ -55,6 +78,8 @@ USER starpunk
# Expose application port
EXPOSE 8000
# MCP sidecar (same image, run with: uvicorn starpunk.mcp_server.app:app --host 0.0.0.0 --port 8001)
EXPOSE 8001
# Health check
# Uses httpx (already in requirements) to verify app is responding
+1 -1
View File
@@ -1,6 +1,6 @@
MIT License
Copyright (c) 2024 [Your Name]
Copyright (c) 2025-2026 Phil Skentelbery
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
+103 -175
View File
@@ -1,199 +1,127 @@
# StarPunk
A minimal, self-hosted IndieWeb CMS for publishing notes with RSS syndication.
**Current Version**: 1.1.0
## Versioning
StarPunk follows [Semantic Versioning 2.0.0](https://semver.org/):
- Version format: `MAJOR.MINOR.PATCH`
- Current: `1.1.0` (stable release)
- Check version: `python -c "from starpunk import __version__; print(__version__)"`
- See changes: [CHANGELOG.md](CHANGELOG.md)
- Versioning strategy: [docs/standards/versioning-strategy.md](docs/standards/versioning-strategy.md)
## Philosophy
A minimal, self-hosted, single-user IndieWeb site for publishing notes.
Flask, SQLite and markdown files; one container.
"Every line of code must justify its existence. When in doubt, leave it out."
StarPunk is designed for a single user who wants to:
- Publish short notes to their personal website
- Own their content (notes stored as portable markdown files)
- Syndicate via RSS
- Support IndieWeb standards (Micropub, IndieAuth)
- Run on minimal resources
The current version is in [`starpunk/__init__.py`](starpunk/__init__.py);
what changed in each release is in [CHANGELOG.md](CHANGELOG.md).
## Features
- **File-based storage**: Notes are markdown files, owned by you
- **IndieAuth authentication**: Use your own website as identity
- **Micropub support**: Full W3C Micropub specification compliance
- **RSS feed**: Automatic syndication
- **No database lock-in**: SQLite for metadata, files for content
- **Self-hostable**: Run on your own server
- **Minimal dependencies**: 6 core dependencies, no build tools
**Publishing**
- Notes in markdown, stored as plain files with metadata in SQLite
- Admin editor with drafts, custom slugs, tags and photo uploads
(resized variants, HEIC/MPO converted to JPEG, captions)
- Post types: notes, titled posts, replies, likes, reposts and bookmarks
- Full-text search (SQLite FTS5) and tag pages
## Requirements
**IndieWeb**
- [Micropub](https://www.w3.org/TR/micropub/): create, update, delete and
undelete; media endpoint; `q=config`, `source`, `syndicate-to`, `category`
- Built-in [IndieAuth](https://indieauth.spec.indieweb.org/) server for
Micropub apps (consent screen, tokens, revocation, Connected Apps page)
- Admin login through any OpenID Connect provider (e.g. Pocket ID)
- [MCP](https://modelcontextprotocol.io/) server at `/mcp` so AI assistants
(claude.ai connectors, Claude Code…) can search, read and write notes
after you approve them, run as an optional sidecar (ADR-075)
- [Webmention](https://www.w3.org/TR/webmention/): receiving (verified,
moderated, rate-limited) and sending; replies, quotes and mentions shown
on each note, with likes, reposts and replies as a line of counts with
hover cards and `<indie-action>` web actions
- [WebSub](https://www.w3.org/TR/websub/) with a built-in hub, so readers get
new posts within seconds
- [Microformats2](https://microformats.org/wiki/h-entry): h-entry, h-feed and
an h-card discovered from your own site
- Python 3.11 or higher
- 500MB disk space
- Linux, macOS, or Windows with WSL2
**Syndication**
- Feeds: RSS 2.0 (with Media RSS), Atom 1.0, JSON Feed 1.1, content
negotiation at `/feed`, OPML export; tag pages and the likes page are
h-feeds readers can follow
- POSSE crossposting to GoToSocial/Mastodon, Bluesky and Threads, with
`u-syndication` links back; edits and deletes follow, native likes and
boosts where the target supports them
- Backfeed: likes, reposts and replies to crossposts on GoToSocial and
Bluesky (and Bluesky quotes), and replies on Threads (with its like and
repost counts), come back as responses (`BACKFEED_ENABLED`, ADR-073)
## Quick Start
**Operations**
- Migrations applied automatically at startup
- `/health` endpoint, `/admin/health` diagnostics, structured logs
- Notifications of new responses and crossposting problems through
Apprise (Gotify, ntfy, email…; `NOTIFY_URLS`, ADR-074)
- Fetches of user-supplied URLs are limited to public addresses, size and
time
## Quick start
### Container
```bash
# Clone repository
git clone https://github.com/YOUR_USERNAME/starpunk.git
cd starpunk
# Install uv (package manager)
curl -LsSf https://astral.sh/uv/install.sh | sh
# Create virtual environment
uv venv .venv --python 3.11
# Install dependencies
uv pip install -r requirements.txt
# Configure
cp .env.example .env
# Edit .env and set ADMIN_ME and SESSION_SECRET
# Initialize database
mkdir -p data/notes
.venv/bin/python -c "from starpunk.database import init_db; init_db()"
# Note: Database also auto-initializes on first run if not present
# Run development server
.venv/bin/flask --app app.py run --debug
# Visit http://localhost:5000
git clone https://git.thesatelliteoflove.com/phil/StarPunk.git && cd StarPunk
cp .env.example .env # set SITE_URL, ADMIN_ME, SESSION_SECRET, OIDC_*
docker build -t starpunk -f Containerfile .
docker run -d --name starpunk -p 127.0.0.1:8000:8000 --env-file .env \
-v ./starpunk-data:/data starpunk
```
The container runs as UID 1000, so the data directory must be writable by
it. Release images are published by CI as
`git.thesatelliteoflove.com/phil/starpunk:vX.Y.Z`. Put a reverse proxy
with HTTPS in front (see `Caddyfile.example` or `nginx.conf.example`) and
set `TRUSTED_PROXY=true`. Register `<SITE_URL>/auth/callback` as the
redirect URI with your OIDC provider. Details:
[docs/architecture/deployment.md](docs/architecture/deployment.md).
### Local development
```bash
uv venv --python 3.14
uv pip install -r requirements-dev.txt
cp .env.example .env # set SESSION_SECRET; add DEV_MODE=true and DEV_ADMIN_ME
uv run flask --app app.py run --debug # http://localhost:5000
SESSION_SECRET=test uv run pytest tests/test_<area>.py # tests for what you changed; CI runs the full suite
```
With `DEV_MODE=true`, `/dev/login` logs you in without OIDC. Never enable
it on a public server. Full guide:
[docs/standards/development-setup.md](docs/standards/development-setup.md).
## Configuration
All configuration is in the `.env` file. Required settings:
All settings are environment variables (or a `.env` file), each listed with
its default in [`.env.example`](.env.example). Required in production:
`SESSION_SECRET`, `SITE_URL`, `ADMIN_ME`, `OIDC_ISSUER`, `OIDC_CLIENT_ID`,
`OIDC_CLIENT_SECRET` and `OIDC_ADMIN_SUB`. Crossposting is enabled by
setting a target's credentials. Backfeed is off unless
`BACKFEED_ENABLED=true`; notifications start when `NOTIFY_URLS` is set.
- `ADMIN_ME` - Your IndieWeb identity URL (e.g., https://yoursite.com)
- `SESSION_SECRET` - Random secret key (generate with `python3 -c "import secrets; print(secrets.token_hex(32))"`)
- `SITE_URL` - Public URL of your site
## Data and backups
See `.env.example` for all options.
Everything lives in the data directory (`/data` in the container): the
SQLite database, note files in `notes/`, uploaded media, saved response
avatars in `avatars/`, the Threads token (`threads_token.json`) and logs.
Back it up regularly, and always before upgrading to a release with a migration.
## Project Structure
## Documentation
```
starpunk/
├── app.py # Application entry point
├── starpunk/ # Application code
├── data/ # Your notes and database (gitignored)
│ ├── notes/ # Markdown files
│ └── starpunk.db # SQLite database
├── static/ # CSS and JavaScript
├── templates/ # HTML templates
└── tests/ # Test suite
```
## Usage
### Publishing Notes
**Via Web Interface**:
1. Navigate to `/admin`
2. Login with your IndieWeb identity
3. Create notes in markdown
**Via Micropub Client**:
1. Configure client with your site URL
2. Authenticate via IndieAuth
3. Publish from any Micropub-compatible app
### Backing Up Your Data
Your notes are stored as plain markdown files in `data/notes/`. Back up this directory:
```bash
# Simple backup
tar -czf backup.tar.gz data/
# Or use rsync
rsync -av data/ /backup/starpunk/
```
## Development
See [docs/standards/development-setup.md](docs/standards/development-setup.md) for detailed setup.
```bash
# Install dev dependencies
uv pip install -r requirements-dev.txt
# Run tests
.venv/bin/pytest
# Format code
.venv/bin/black starpunk/ tests/
# Lint
.venv/bin/flake8 starpunk/ tests/
```
## Architecture
StarPunk uses a hybrid storage approach:
- **Notes content**: Markdown files (portable, human-readable)
- **Metadata**: SQLite database (fast queries)
This gives you both portability AND performance.
See [docs/architecture/](docs/architecture/) for complete documentation.
## IndieWeb Compliance
StarPunk implements:
- [Micropub](https://micropub.spec.indieweb.org/) - Publishing API
- [IndieAuth](https://www.w3.org/TR/indieauth/) - Authentication
- [Microformats2](http://microformats.org/) - Semantic HTML markup
- [RSS 2.0](https://www.rssboard.org/rss-specification) - Feed syndication
## Deployment
### Production Setup
```bash
# Install gunicorn
uv pip install gunicorn
# Run with gunicorn
.venv/bin/gunicorn -w 4 -b 127.0.0.1:8000 app:app
# Configure nginx/Caddy for HTTPS
# Set up systemd for process management
# Enable regular backups of data/ directory
```
See [docs/standards/deployment-standards.md](docs/standards/deployment-standards.md) for details.
| Topic | Where |
|---|---|
| How the docs are organized | [docs/DOCUMENTATION.md](docs/DOCUMENTATION.md) |
| System overview | [docs/architecture/overview.md](docs/architecture/overview.md) |
| Technology stack | [docs/architecture/technology-stack.md](docs/architecture/technology-stack.md) |
| Database and migrations | [docs/architecture/database.md](docs/architecture/database.md) |
| Deployment | [docs/architecture/deployment.md](docs/architecture/deployment.md) |
| Authentication (OIDC, IndieAuth) | [docs/architecture/simplified-auth-architecture.md](docs/architecture/simplified-auth-architecture.md) |
| IndieWeb protocols | [docs/architecture/indieweb-protocols.md](docs/architecture/indieweb-protocols.md) |
| Feeds and crossposting | [docs/architecture/syndication-architecture.md](docs/architecture/syndication-architecture.md) |
| Decisions (ADRs) | [docs/decisions/INDEX.md](docs/decisions/INDEX.md) |
| Per-release designs and reports | [docs/design/](docs/design/) |
| Backlog | [docs/projectplan/BACKLOG.md](docs/projectplan/BACKLOG.md) |
| Development standards and release process | [docs/standards/INDEX.md](docs/standards/INDEX.md) |
## License
MIT License - see LICENSE file
## Credits
Built with:
- [Flask](https://flask.palletsprojects.com/) - Web framework
- [python-markdown](https://python-markdown.github.io/) - Markdown processing
- [feedgen](https://feedgen.kiesow.be/) - RSS generation
- [httpx](https://www.python-httpx.org/) - HTTP client
- [IndieLogin](https://indielogin.com/) - Authentication service
## Contributing
This is a personal project optimized for single-user use. If you want additional features, consider forking!
## Support
- Documentation: [docs/](docs/)
- Issues: GitHub Issues
- IndieWeb: [indieweb.org](https://indieweb.org/)
[MIT](LICENSE): use, change and share it for anything, including
commercially, as long as the copyright and license notice stay with it.
+2 -7
View File
@@ -11,12 +11,11 @@
# docker compose logs -f
# docker compose down
version: '3.8'
services:
starpunk:
# Container configuration
image: starpunk:0.6.0
# Built locally from this checkout (release images: see README)
image: starpunk:local
container_name: starpunk
# Build configuration
@@ -49,10 +48,6 @@ services:
- NOTES_PATH=/data/notes
- DATABASE_PATH=/data/starpunk.db
# Application metadata
- VERSION=0.6.0
- ENVIRONMENT=production
# Volume mounts for persistent data
# All application data stored in ./container-data on host
volumes:
-103
View File
@@ -1,103 +0,0 @@
"""
Development authentication module for StarPunk
WARNING: This module provides a development-only authentication mechanism
that bypasses IndieLogin. It should NEVER be enabled in production.
This module is separate from production auth (starpunk/auth.py) to maintain
clear architectural boundaries and enable easy security audits.
Security measures:
- Only active when DEV_MODE=true
- Returns 404 if DEV_MODE=false
- Requires DEV_ADMIN_ME configuration
- Logs prominent warnings
- Cannot coexist with production SITE_URL
- Visual warnings in UI
Functions:
is_dev_mode: Check if development mode is enabled
validate_dev_config: Validate development configuration
create_dev_session: Create session without authentication
"""
from flask import current_app
from starpunk.auth import create_session
def is_dev_mode() -> bool:
"""
Check if development mode is enabled
Returns:
True if DEV_MODE is enabled, False otherwise
Example:
>>> from starpunk.dev_auth import is_dev_mode
>>> if is_dev_mode():
... print("Development mode active")
"""
return current_app.config.get("DEV_MODE", False)
def validate_dev_config() -> None:
"""
Validate development mode configuration
Checks that DEV_MODE configuration is valid and safe:
- DEV_ADMIN_ME must be set if DEV_MODE is true
- Warns if DEV_MODE is enabled with production-like SITE_URL
Raises:
ValueError: If DEV_MODE is true but DEV_ADMIN_ME is not set
Logs:
WARNING: If DEV_MODE is enabled with HTTPS SITE_URL
"""
dev_mode = current_app.config.get("DEV_MODE", False)
if dev_mode:
# Require DEV_ADMIN_ME
dev_admin_me = current_app.config.get("DEV_ADMIN_ME")
if not dev_admin_me:
raise ValueError("DEV_MODE=true requires DEV_ADMIN_ME to be set")
# Warn if production-like configuration detected
site_url = current_app.config.get("SITE_URL", "")
if site_url.startswith("https://"):
current_app.logger.warning(
"WARNING: DEV_MODE is enabled with production SITE_URL. "
"This is likely a misconfiguration. "
"DEV_MODE should only be used in local development."
)
def create_dev_session(me: str) -> str:
"""
Create development session without authentication
WARNING: This bypasses IndieLogin authentication entirely.
Only use in development environments.
Args:
me: User identity URL (from DEV_ADMIN_ME config)
Returns:
Session token (same format as production sessions)
Logs:
WARNING: Logs that dev session was created without authentication
Example:
>>> token = create_dev_session("https://example.com")
>>> # Session created without IndieLogin verification
"""
current_app.logger.warning(
f"DEV MODE: Creating session for {me} without authentication. "
f"This should NEVER happen in production!"
)
# Use production session creation (same session format)
# This ensures dev sessions work identically to production
return create_session(me)
+57
View File
@@ -0,0 +1,57 @@
# PURPOSE
This document describes how documentation in this folder should be organized and supersedes any other instructions.
# FOLDERS
## ARCHITECTURE
The architecture folder should contain documentation reflecting the current design of the system and should be updated at the end of each release to ensure it is current.
## DECISIONS
This folder contains any architectural decisions, documented as ADRs.
- Format: `ADR-NNN-brief-title.md` (numbered sequentially)
- Create an ADR when making architectural decisions, choosing between technical approaches, or establishing patterns
## DESIGN
This folder is used by the architect to document implementation designs to be handed off to the developer. These designs should be sorted into subfolders reflecting the semantic version number of the release in question (e.g., `v1.0.0/`, `v1.1.1/`).
All developer interaction belongs in the appropriate version subfolder:
- Implementation designs and specifications
- Developer questions to the architect
- Architect responses
- Implementation reports (format: `YYYY-MM-DD-description.md`)
- Implementation reviews
## PROJECTPLAN
This folder contains documents relating to the future state of the project. There should be a single BACKLOG.md file that lists future features by priority as well as bugs (which are assumed to be high priority). Items in this file can have one of the following priorities:
- Critical - Items that break existing functionality
- High
- Medium
- Low
In addition to the backlog file each version should have a folder named for its semantic version with a RELEASE.md file which lists the features and bugs to be addressed in that release.
## STANDARDS
Includes any standards written by the architect that the developer needs to reference during development. Any deprecated standards should be moved to the DEPRECATED subfolder when appropriate.
# WHERE TO FIND DOCUMENTATION
- **Before implementing a feature**: Check `decisions/` for relevant ADRs and `design/{version}/` for specifications
- **Understanding system architecture**: Start with `architecture/`
- **Coding guidelines**: See `standards/`
- **Past implementation context**: Review `design/{version}/` for similar work
- **Project roadmap and scope**: Refer to `projectplan/`
# KEY REFERENCES
- **Architecture**: `architecture/`
- **Coding Standards**: `standards/python-coding-standards.md`
- **Testing**: `standards/testing-checklist.md`
- **Project Backlog**: `projectplan/BACKLOG.md`
+31 -71
View File
@@ -1,82 +1,42 @@
# Architecture Documentation Index
This directory contains architectural documentation, system design overviews, component diagrams, and architectural patterns for StarPunk CMS.
This directory describes the **current** design of StarPunk (v2.14.0). Per
[../DOCUMENTATION.md](../DOCUMENTATION.md) it is updated at the end of each
release. Decisions and their history are in [../decisions/](../decisions/);
per-release designs and reports are in [../design/](../design/). The
original v0.x technology choices are kept in
[../design/v0.x/technology-stack-legacy.md](../design/v0.x/technology-stack-legacy.md).
## Core Architecture
## Documents
### System Overview
- **[overview.md](overview.md)** - Complete system architecture and design principles
- **[technology-stack.md](technology-stack.md)** - Current technology stack and dependencies
- **[technology-stack-legacy.md](technology-stack-legacy.md)** - Historical technology decisions
| Document | Description |
|----------|-------------|
| [overview.md](overview.md) | System overview: purpose, request flow, modules and routes, background workers, storage, auth model, URL surface, caching, security posture |
| [technology-stack.md](technology-stack.md) | Current runtime, Python dependencies, standards implemented, testing and CI |
| [database.md](database.md) | SQLite schema after migrations 001–022, migration runner and its lock, connection pool |
| [deployment.md](deployment.md) | Container image stages, CI workflow, release process, `/data` layout, reverse proxy, all environment variables |
| [simplified-auth-architecture.md](simplified-auth-architecture.md) | Admin login via OpenID Connect and the built-in IndieAuth server that issues Micropub tokens (ADR-064) |
| [indieweb-protocols.md](indieweb-protocols.md) | How Micropub, IndieAuth, Webmention, microformats2, WebSub and the feeds are implemented (including backfeed responses, web actions and notifications), with compliance status and known gaps |
| [mcp.md](mcp.md) | The MCP server for AI clients: sidecar, authorization flow, tools (ADR-075) |
| [syndication-architecture.md](syndication-architecture.md) | Crossposting (POSSE) to GoToSocial, Bluesky and Threads: syndications table, worker, reconcile rules, backfeed of responses (ADR-073) |
### Feature-Specific Architecture
## Where to start
#### IndieAuth & Authentication
- **[indieauth-assessment.md](indieauth-assessment.md)** - Assessment of IndieAuth implementation
- **[indieauth-client-diagnosis.md](indieauth-client-diagnosis.md)** - IndieAuth client diagnostic analysis
- **[indieauth-endpoint-discovery.md](indieauth-endpoint-discovery.md)** - Endpoint discovery architecture
- **[indieauth-identity-page.md](indieauth-identity-page.md)** - Identity page architecture
- **[indieauth-questions-answered.md](indieauth-questions-answered.md)** - Architectural Q&A for IndieAuth
- **[indieauth-removal-architectural-review.md](indieauth-removal-architectural-review.md)** - Review of custom IndieAuth removal
- **[indieauth-removal-implementation-guide.md](indieauth-removal-implementation-guide.md)** - Implementation guide for removal
- **[indieauth-removal-phases.md](indieauth-removal-phases.md)** - Phased removal approach
- **[indieauth-removal-plan.md](indieauth-removal-plan.md)** - Overall removal plan
- **[indieauth-token-verification-diagnosis.md](indieauth-token-verification-diagnosis.md)** - Token verification diagnostic analysis
- **[simplified-auth-architecture.md](simplified-auth-architecture.md)** - Simplified authentication architecture
- **[endpoint-discovery-answers.md](endpoint-discovery-answers.md)** - Endpoint discovery implementation Q&A
- **New to the code**: overview.md, then technology-stack.md.
- **Changing the schema**: database.md (migrations must be safe inside one
transaction).
- **Releasing or operating the site**: deployment.md.
- **Working on auth, Micropub, Webmention or WebSub**:
simplified-auth-architecture.md and indieweb-protocols.md, plus the ADRs
they link to.
#### Database & Migrations
- **[database-migration-architecture.md](database-migration-architecture.md)** - Database migration system architecture
- **[migration-fix-quick-reference.md](migration-fix-quick-reference.md)** - Quick reference for migration fixes
- **[migration-race-condition-answers.md](migration-race-condition-answers.md)** - Race condition resolution Q&A
## Related documentation
#### Syndication
- **[syndication-architecture.md](syndication-architecture.md)** - RSS feed and syndication architecture
## Version-Specific Architecture
### v1.0.0
- **[v1.0.0-release-validation.md](v1.0.0-release-validation.md)** - Release validation architecture
### v1.1.0
- **[v1.1.0-feature-architecture.md](v1.1.0-feature-architecture.md)** - Feature architecture for v1.1.0
- **[v1.1.0-implementation-decisions.md](v1.1.0-implementation-decisions.md)** - Implementation decisions
- **[v1.1.0-search-ui-validation.md](v1.1.0-search-ui-validation.md)** - Search UI validation
- **[v1.1.0-validation-report.md](v1.1.0-validation-report.md)** - Overall validation report
### v1.1.1
- **[v1.1.1-architecture-overview.md](v1.1.1-architecture-overview.md)** - Architecture overview for v1.1.1
## Phase Documentation
- **[phase1-completion-guide.md](phase1-completion-guide.md)** - Phase 1 completion guide
- **[phase-5-validation-report.md](phase-5-validation-report.md)** - Phase 5 validation report
## Review Documentation
- **[review-v1.0.0-rc.5.md](review-v1.0.0-rc.5.md)** - Architectural review of v1.0.0-rc.5
## How to Use This Documentation
### For New Developers
1. Start with **overview.md** to understand the system
2. Review **technology-stack.md** for current technologies
3. Read feature-specific architecture docs relevant to your work
### For Architects
1. Review version-specific architecture for historical context
2. Consult feature-specific docs when making changes
3. Update relevant docs when architecture changes
### For Contributors
1. Read **overview.md** for system understanding
2. Consult specific architecture docs for areas you're working on
3. Follow patterns documented in architecture files
## Related Documentation
- **[../decisions/](../decisions/)** - Architectural Decision Records (ADRs)
- **[../design/](../design/)** - Detailed design documents
- **[../standards/](../standards/)** - Coding standards and conventions
- [../decisions/](../decisions/): Architectural Decision Records (ADRs)
- [../design/](../design/): designs, questions and reports per version
- [../standards/](../standards/): coding, testing and versioning standards
- [../projectplan/](../projectplan/): backlog and release plans
---
**Last Updated**: 2025-11-25
**Maintained By**: Documentation Manager Agent
**Last Updated**: 2026-11-10 (v2.14.0)
+256
View File
@@ -0,0 +1,256 @@
# Database Architecture
**Version**: 2.14.0 (2026-11-10)
**Status**: Current. Verified against a database created by `create_app()`
with migrations 001–022.
StarPunk uses one SQLite file, `starpunk.db` under `DATA_PATH` (`/data` in
the container), in WAL mode. Note *content* is not in the database: it lives
in Markdown files (`notes/YYYY/MM/<slug>.md`) and the `notes` table points to
them. See [overview.md](overview.md) for how the rest of the system uses it.
## How the schema is built
On every start, `init_db()` (`starpunk/database/init.py`) does two things:
1. Runs `INITIAL_SCHEMA_SQL` (`starpunk/database/schema.py`). This is the
frozen v1.0.0 baseline, written with `CREATE … IF NOT EXISTS`: `notes`,
`sessions`, `tokens`, `authorization_codes`, `auth_state` and their
indexes. It must not be changed; every later change is a migration.
2. Runs `run_migrations()` (`starpunk/migrations.py`) over the SQL files in
`migrations/`, in file-name order.
### Migration runner
- Applied migrations are recorded in `schema_migrations`
(`migration_name`, `applied_at`).
- **Locking**: all four gunicorn processes run `create_app()` at the same
time. Each opens a direct (unpooled) connection with a 30-second busy
timeout and runs `BEGIN IMMEDIATE`, which takes SQLite's write lock. The
whole run (creating `schema_migrations`, checking what is pending, applying
every pending migration and recording it) happens in that one transaction
and is committed at the end, so a migration is applied once and atomically.
Statements are executed one by one with `conn.execute()` (split with
`sqlite3.complete_statement`, so trigger bodies stay whole), never with
`executescript()`, which would commit and drop the lock part-way (the cause
of the v2.8.0 deploy failure fixed in v2.8.1).
- A process that finds the database locked retries with exponential backoff
(up to 10 attempts, 120 seconds in total), then finds everything already
applied and returns.
- A failing migration rolls the whole run back and raises `MigrationError`,
which stops the app from starting.
- **Fresh databases**: when `schema_migrations` is empty, the runner checks
each of the early migrations (001–003) against the actual schema and only
records those whose effect is already present; migration 002 is reduced to
creating its indexes. All later migrations run normally.
Background: [ADR-031](../decisions/ADR-031-database-migration-system-redesign.md),
[ADR-032](../decisions/ADR-032-initial-schema-sql-implementation.md),
[ADR-033](../decisions/ADR-033-database-migration-redesign.md),
[ADR-037](../decisions/ADR-037-migration-race-condition-fix.md),
[ADR-041](../decisions/ADR-041-database-migration-conflict-resolution.md).
### Migrations
| File | Version | Change |
|------|---------|--------|
| `001_add_code_verifier_to_auth_state.sql` | 0.8 | `auth_state.code_verifier` (removed again by 003) |
| `002_secure_tokens_and_authorization_codes.sql` | 1.0.0-rc | Hashed `tokens` and `authorization_codes` (dropped by 004) |
| `003_remove_code_verifier_from_auth_state.sql` | 1.0 | Drops `auth_state.code_verifier` |
| `004_drop_token_tables.sql` | 1.0 | Drops `tokens` and `authorization_codes` |
| `005_add_fts5_search.sql` | 1.1 | `notes_fts` (FTS5) and the `notes_fts_delete` trigger |
| `006_add_author_profile.sql` | 1.2 | `author_profile` (h-card cache) |
| `007_add_media_support.sql` | 1.2 | `media`, `note_media` |
| `008_add_tags.sql` | 1.3 | `tags`, `note_tags` |
| `009_add_media_variants.sql` | 1.4 | `media_variants` |
| `010_add_webmentions.sql` | 1.6 | `notes.in_reply_to`, `webmentions`, `webmention_queue`, `reply_contexts` |
| `011_oidc_login_and_indieauth_server.sql` | 2.0 | `oidc_login_state`, `indieauth_codes`, `indieauth_tokens` |
| `012_add_syndications.sql` | 2.1 | `syndications` |
| `013_add_likes_reposts.sql` | 2.5 | `notes.like_of`, `notes.repost_of` |
| `014_webmention_source_details.sql` | 2.8 | `webmentions.source_post_url`, `webmentions.source_published` |
| `015_bookmarks_and_content_format.sql` | 2.9 | `notes.bookmark_of`, `notes.content_format` |
| `016_websub.sql` | 2.10.0 | `websub_subscriptions`, `websub_intents`, `websub_deliveries`, `site_state` |
| `017_websub_hardening.sql` | 2.10.1 | `websub_intents.request_id`, `requested_topic`, `callback_ip`; `websub_subscriptions.callback_ip` |
| `018_foreign_key_cleanup.sql` | 2.11.0 | Removes rows orphaned while foreign keys weren't enforced |
| `019_shared_rate_limits.sql` | 2.11.0 | `rate_limits` (counters shared by all processes) |
| `020_backfeed.sql` | 2.13.0 | Rebuilds `webmentions` adding `origin`, `remote_id`, the `quote` type and an index on (`target_note_id`, `origin`); `syndications.backfeed_next_at`, `backfeed_checked_at`, `backfeed_claimed_until`, `backfeed_attempts`, `backfeed_error`, `backfeed_backfill`, `likes_count`, `reposts_count`, `quotes_count` |
| `021_notifications.sql` | 2.14.0 | `webmentions.notified_at`; existing rows are marked as notified |
| `022_indieauth_resource.sql` | 2.15.0 | `indieauth_codes.resource` and `indieauth_tokens.resource` (RFC 8707, MCP tokens; ADR-075) |
New migrations are numbered `NNN_description.sql`, must be safe to run inside
one transaction, and should use `IF NOT EXISTS` where SQLite allows it.
## Tables
### Content
**`notes`**: one row per note; the content is in the file at `file_path`.
| Column | Purpose |
|--------|---------|
| `id`, `slug` (unique), `file_path` (unique) | Identity and location of the Markdown file (relative to `DATA_PATH`) |
| `published` | 0 = draft, 1 = public |
| `created_at`, `updated_at` | Timestamps (UTC) |
| `deleted_at` | Set on soft delete (file moved to `.trash/`); cleared by undelete |
| `content_hash` | SHA-256 of the file content |
| `in_reply_to`, `like_of`, `repost_of`, `bookmark_of` | URL the note responds to; set for replies, likes, reposts, bookmarks |
| `content_format` | `'html'` when a Micropub client sent HTML (so `q=source` returns HTML); NULL means Markdown |
Indexes on `created_at DESC`, `published`, `slug`, `deleted_at`, `like_of`.
**`notes_fts`**: FTS5 virtual table (`slug` unindexed, `title`, `content`;
`porter unicode61` tokenizer), with its shadow tables `notes_fts_config`,
`_content`, `_data`, `_docsize`, `_idx`. Rows use the note `id` as `rowid`.
Inserts and updates are done by `starpunk/search.py` (the database cannot
read note files); the trigger `notes_fts_delete` removes the row when a note
row is deleted. The index is rebuilt at start-up if it is empty.
**`tags`** (`name` normalised and unique, `display_name` as written) and
**`note_tags`** (`note_id`, `tag_id`, unique pair, cascading deletes).
**`media`**: an uploaded image: original `filename`, `stored_filename`
(UUID), `path` (`YYYY/MM/<uuid>.<ext>` under `media/`), `mime_type`, `size`,
`width`, `height`, `uploaded_at`.
**`media_variants`**: resized copies per image (`variant_type` one of
`thumb`, `small`, `medium`, `large`, `original`; `path`, dimensions, size);
unique per (`media_id`, `variant_type`), deleted with the image.
**`note_media`**: attaches media to notes with `display_order` and
`caption` (alt text); unique per (`note_id`, `media_id`).
### Authentication
**`sessions`**: admin sessions. `session_token_hash` (SHA-256 of the cookie
value), `me`, `created_at`, `expires_at`, `last_used_at`, `user_agent`,
`ip_address`.
**`oidc_login_state`**: one row per OIDC login in progress: `state`
(primary key), `nonce`, `code_verifier` (PKCE), `next_url`, `expires_at`.
Consumed once on callback.
**`indieauth_codes`**: authorization codes issued to Micropub and MCP clients:
`code_hash`, `client_id`, `redirect_uri`, `scope`, `code_challenge`,
`resource` (`SITE_URL/mcp` or NULL), `expires_at`, `used_at` (single use).
**`indieauth_tokens`**: access tokens: `token_hash`, `client_id`, `scope`,
`resource` (`SITE_URL/mcp` for MCP tokens, NULL for Micropub), `created_at`, `expires_at`, `last_used_at`, `revoked_at`. Listed and revoked
in `/admin/tokens`.
**Legacy tables, unused**: `auth_state` (v1 IndieLogin login state), `tokens`
and `authorization_codes` (v1 token tables). No code reads or writes them.
Migration 004 drops `tokens` and `authorization_codes`, but
`INITIAL_SCHEMA_SQL` recreates them (empty) on the next start, so a running
database normally contains them. They are harmless; v2 auth uses the tables
above.
### IndieWeb interactions
**`webmentions`**: received webmentions and backfed silo responses
(v2.13.0), one per (`source_url`, `target_url`).
| Column | Purpose |
|--------|---------|
| `target_note_id` | The note mentioned (SET NULL if the note is deleted) |
| `mention_type` | `reply`, `like`, `repost`, `bookmark`, `mention`, `quote` |
| `author_name`, `author_url`, `author_photo`, `content_snippet` | From the source's mf2 (content sanitised) |
| `source_post_url`, `source_published` | The source entry's own `u-url` and `dt-published` |
| `validation_status` | `pending`, `valid`, `invalid`, `error` (set by the worker) |
| `moderation_status` | `pending`, `approved`, `hidden`, `deleted` (auto-approved when `WEBMENTION_AUTO_APPROVE`). `deleted` is a tombstone: deleting in the admin keeps the row, so a re-sent webmention or the next backfeed check doesn't bring it back; re-validation keeps `hidden` and `deleted` |
| `origin` | `webmention`, or the silo it was backfed from (`gotosocial`, `bluesky`, `threads`) |
| `remote_id` | Silo id of a backfed response (or of the person, for likes and reposts) |
| `notified_at` | When the owner was notified of it (v2.14.0); NULL means not yet |
| `validation_error`, `validated_at`, `received_at`, `updated_at` | Bookkeeping |
**`webmention_queue`**: outgoing webmentions, one per (`source_url`,
`target_url`): `source_note_id`, discovered `endpoint_url`, `status`
(`pending`, `sending`, `sent`, `failed`, `no_endpoint`), `retry_count`,
`next_retry_at`, `last_error`, `started_at`, `sent_at`.
**`reply_contexts`**: cache of the posts that replies (and other responses)
point to: `url` (unique), `title`, `content_snippet`, `published_at`, author
fields, `site_name`, `fetch_status` (`pending`, `fetched`, `error`),
`fetch_error`, `fetched_at`.
**`author_profile`**: cache of the owner's h-card discovered from `ADMIN_ME`:
`me`, `name`, `photo`, `url`, `note`, `rel_me_links` (JSON), `discovered_at`,
`cached_until` (24 hours).
### Crossposting
**`syndications`**: one row per (`note_id`, `target`), where `target` is a
configured target id (GoToSocial, Bluesky, Threads). `enabled` says whether
the note should be on that target; `remote_id` and `remote_url` identify the
remote post. `state` (`pending`, `synced`, `failed`), `version` (bumped on
every note change, so an edit made during a send is not lost), `attempts`,
`last_error`, `next_attempt_at` and `claimed_until` (a 5-minute lease that
lets any gunicorn process take the row exclusively) drive the worker.
Backfeed (v2.13.0) adds `backfeed_next_at`, `backfeed_checked_at`,
`backfeed_claimed_until` (its own lease), `backfeed_attempts`,
`backfeed_error` and `backfeed_backfill` (set while a backfill is due), and
`likes_count`, `reposts_count`, `quotes_count`: silo-only counts (all of
Threads; elsewhere people who opted out). See
[syndication-architecture.md](syndication-architecture.md).
### WebSub
**`websub_subscriptions`**: verified subscriptions to this site's topics:
`topic`, `callback` (unique pair), `secret`, `lease_seconds`, `expires_at`,
`verified_at`, `last_success_at`, `last_error`, `callback_ip` (the address
the callback resolved to, for per-address limits).
**`websub_intents`**: subscribe and unsubscribe requests waiting for
verification of intent: `topic`, `callback` (unique pair), `mode`, `secret`,
`lease_seconds`, `attempts`, `next_attempt_at`, `request_id` (so verifying an
older request cannot consume a newer one), `requested_topic` (the topic as
sent, echoed in verification), `callback_ip`.
**`websub_deliveries`**: content waiting to be pushed: one row per
subscription (built-in hub) or one per topic with `subscription_id` NULL (a
publish ping to an external hub); partial unique indexes keep at most one
pending row of each kind. `attempts`, `next_attempt_at`, `last_error`.
**`site_state`**: key/value pairs. `content_version` is bumped whenever a
public note changes; every gunicorn process compares it to decide whether
its cached feed data is stale. Backfeed keeps `backfeed_backfill_last` (spacing
backfill checks across processes), `backfeed_backfill_started` and
`backfeed_backfill_finished`; notifications keep `notify_last_error` (shown
in the admin, cleared by a successful send).
### Bookkeeping
**`schema_migrations`**: applied migrations (see above).
**`rate_limits`** (v2.11.0): request counts for `/webmention` and `/websub`,
one row per key (`<limiter>:<ip or domain>`) and time window, so every
server process enforces one shared limit. Old windows are pruned as it goes.
## Connections
- **Request connections**: `starpunk/database/pool.py`. `init_pool()` creates
one `ConnectionPool` per process (up to 5 idle connections, 10-second busy
timeout; `DB_POOL_SIZE` and `DB_TIMEOUT` can be set only through the
`create_app()` config dict). Connections use `sqlite3.Row` rows,
`check_same_thread=False`, `PRAGMA journal_mode=WAL` and (since v2.11.0)
`PRAGMA foreign_keys=ON`, so `ON DELETE CASCADE`/`SET NULL` run. `get_db()`
stores one connection in `flask.g` per request or app context; on teardown
it goes back to the pool, rolled back first if a transaction was left
open.
- **Direct connections**: the migration runner, the FTS start-up check, the
webmention worker and sender, the notification sweep (`notifications.py`),
avatar upkeep (`avatars.py`), reply-context fetching and search helpers
open their own `sqlite3.connect(db_path)` connections.
- WAL mode lets readers carry on while one writer writes. The four gunicorn
processes and the worker threads share the one file; writes are short, so
contention is low.
## Backups
The application does not back itself up. A consistent copy needs the
database (use SQLite's online backup, e.g. `sqlite3 starpunk.db ".backup …"`,
because of WAL) together with `notes/` and `media/`. `avatars/` is optional:
missing avatars are fetched again. See [deployment.md](deployment.md).
---
**Last Updated**: 2026-11-10 (v2.14.0)
+283
View File
@@ -0,0 +1,283 @@
# Deployment Architecture
**Version**: 2.14.0 (2026-11-10)
**Status**: Current
How StarPunk is built, configured and run in production. The production site
is a single container behind Caddy on one host, managed with Ansible. This
document stays generic: hostnames, credentials and inventory live in the
deployment repository, not here.
## Overview
```
git tag vX.Y.Z ──► Gitea Actions ──► test stage (pytest) ──► runtime image
│ push :vX.Y.Z, :latest
▼
Gitea container registry
│ pull (tag set in Ansible)
▼
Internet ──► Caddy (TLS, headers, X-Forwarded-For) ──► container :8000 (gunicorn × 4)
│
/data volume (host)
```
## Container image (`Containerfile`)
A multi-stage build on `python:3.14-slim`; it works with Podman and Docker.
| Stage | Purpose |
|-------|---------|
| `builder` | Copies `uv` from `ghcr.io/astral-sh/uv`, creates `/opt/venv` and installs `requirements.txt` into it |
| `test` | Copies the venv, installs `requirements-test.txt`, copies the source and runs `python -m pytest -q -p no:cacheprovider tests` with test values for `SESSION_SECRET`, `SITE_URL` and `ADMIN_ME`. The stage only builds if the tests pass. |
| runtime (last, the default target) | Creates user `starpunk` (UID/GID 1000) and `/data/notes`, copies the venv and the application to `/app`, sets `DATA_PATH=/data`, `NOTES_PATH=/data/notes`, `DATABASE_PATH=/data/starpunk.db`, runs as `starpunk`, exposes port 8000 |
The runtime image's command is gunicorn:
```
gunicorn --bind 0.0.0.0:8000 --workers 4 --worker-class sync
--worker-tmp-dir /dev/shm --max-requests 1000 --max-requests-jitter 50
--timeout 30 --graceful-timeout 30
--access-logfile - --error-logfile - --log-level info app:app
```
Each of the four worker processes builds its own Flask app, so each runs
migrations (only one applies them, see [database.md](database.md)) and starts
its own background threads (see [overview.md](overview.md#background-workers)).
Workers are recycled after about 1000 requests; a recycled process starts
new threads, and the file locks pass to another process if the holder exits.
**Health check**: the image's `HEALTHCHECK` runs every 30 seconds (3-second
timeout, 10-second start period, 3 retries) and calls
`http://localhost:8000/health` with httpx. `/health` returns
`{"status": "ok", "version": …}` plus webmention-worker statistics, without
authentication.
## CI/CD (`.gitea/workflows/build-container.yml`)
- **Triggers**: pushing a tag matching `vX.Y.Z`, or a manual run.
- **Runner**: a Gitea Actions runner with no Docker daemon. Images are built
by a rootless BuildKit daemon reached over its socket
(`docker buildx create --driver remote`).
- **Steps**:
1. Log in to the Gitea container registry with the repository secret
`REGISTRY_TOKEN` (a token with only the `write:package` scope; the
automatic Actions token is refused by the registry).
2. `docker buildx build --target test`: runs the test suite. A failure
stops the workflow before anything is pushed.
3. Build the runtime image and push it tagged with the git tag
(e.g. `v2.10.1`) and `latest`. The image name is the repository path in
lower case.
4. Remove the builder and log out, even after a failure.
There is no CI on branches or pull requests. Locally, run the tests for the
area you changed; CI runs the full suite on the tag (see
[technology-stack.md](technology-stack.md#testing)).
## Release process
Branching, version bump, changelog, merge and tagging are in
[git-branching-strategy.md](../standards/git-branching-strategy.md#release-process).
Pushing a `vX.Y.Z` tag makes CI test, build and push the image. Then, in the
Ansible-managed deployment, set `starpunk_version` to the new tag and deploy
the `starpunk` service: the container is recreated from the new image and new
migrations run at start-up. Check that `/health` reports the new version.
Rolling back means deploying the previous tag. Migrations are not reversed
automatically, so a release with a migration needs a database backup first
if a rollback might be needed.
## Runtime layout
### Data volume (`/data`)
A host directory is mounted at `/data`, owned by UID 1000. Everything
persistent is there; the image itself holds no state.
| Path | Contents |
|------|----------|
| `starpunk.db` (+ `-wal`, `-shm`) | SQLite database in WAL mode |
| `notes/YYYY/MM/<slug>.md` | Note content |
| `.trash/YYYY/MM/` | Soft-deleted note files (restorable with Micropub undelete) |
| `media/YYYY/MM/` | Uploaded images and their variants; `media/.tmp/` while processing |
| `logs/starpunk.log` | Application log, rotated at 10 MB, 10 files kept |
| `webmention-worker.lock`, `websub-worker.lock` | `flock` files choosing the process that runs each worker |
| `avatars/<sha256>.webp` | Saved copies of response authors' avatars (v2.13.0), served at `/avatars/`; unused ones removed weekly |
| `threads_token.json` (+ `.lock`) | Current Threads access token, refreshed daily (only with Threads configured) |
| `debug/` | Failed uploads, only with `DEBUG_SAVE_FAILED_UPLOADS=true`; cleaned up after `DEBUG_FILE_MAX_AGE_DAYS` |
The application does not make backups. A backup must include the database
(copied with SQLite's online backup because of WAL), `notes/` and `media/`.
Backups are handled by the deployment, outside the container.
### Reverse proxy
Caddy terminates TLS and proxies to the container's port 8000. Publish that
port only where the proxy can reach it (for example on the host's loopback
interface, as `compose.yaml` does), never publicly. Caddy adds security headers
(HSTS, CSP, X-Frame-Options, Referrer-Policy and so on; see
`Caddyfile.example`, and `nginx.conf.example` for nginx) and sets
`X-Forwarded-For`.
Set `TRUSTED_PROXY=true` behind the proxy. With it, StarPunk takes the client
address from the first entry of `X-Forwarded-For`, which the webmention and
WebSub rate limits need; without it, every client appears as the proxy's
address and shares one limit. Do not set it if the container port can be
reached without the proxy, because the header could then be forged.
The MCP sidecar (v2.15.0, ADR-075) is a second container from the same
image, with the same environment and `/data` volume plus
`BACKGROUND_WORKERS_ENABLED=false`, running
`uvicorn starpunk.mcp_server.app:app --host 0.0.0.0 --port 8001`. The proxy
sends `/mcp*` and `/.well-known/oauth-protected-resource*` to it, keeping the
`Host` header; everything else goes to port 8000. Without the sidecar the
site simply has no MCP endpoint. See [mcp.md](mcp.md).
`SITE_URL` must be the public `https://` URL: it is used for absolute links,
feed and WebSub topic URLs, the IndieAuth issuer and redirect URIs, and to
decide whether the session cookie is `Secure`.
## Configuration
Configuration is read from environment variables (and a `.env` file if
present) by `starpunk/config.py` at start-up. In production the variables
come from the Ansible-managed environment, with secrets from its vault.
`.env.example` documents them for local use. Booleans are `true`/`false`
unless noted.
### Required
| Variable | Default | Notes |
|----------|---------|-------|
| `SESSION_SECRET` | none | Required always. Signs sessions; also Flask's `SECRET_KEY` unless `FLASK_SECRET_KEY` is set |
| `SITE_URL` | `http://localhost:5000` | Public URL; a trailing `/` is added |
| `ADMIN_ME` | none | Owner's identity URL (h-card, Micropub `me`). Required unless `DEV_MODE` |
| `OIDC_ISSUER`, `OIDC_CLIENT_ID`, `OIDC_CLIENT_SECRET`, `OIDC_ADMIN_SUB` | empty | Admin login via OIDC. Required unless `DEV_MODE` (or testing) |
### Site and authentication
| Variable | Default | Notes |
|----------|---------|-------|
| `SITE_NAME` | `StarPunk` | |
| `SITE_AUTHOR` | `Unknown` | |
| `SITE_DESCRIPTION` | `A minimal IndieWeb CMS` | |
| `SESSION_LIFETIME` | `30` | Days |
| `FLASK_SECRET_KEY` | `SESSION_SECRET` | |
| `OIDC_PROVIDER_NAME` | `single sign-on` | Label on the login button |
| `INDIEAUTH_TOKEN_LIFETIME_DAYS` | `90` | Lifetime of Micropub access tokens |
| `DEV_MODE` | `false` | Enables `/dev/login`; never in production |
| `DEV_ADMIN_ME` | empty | Required when `DEV_MODE=true` |
| `CSRF_ENABLED` | `true` | CSRF tokens on admin forms, logout and IndieAuth consent; leave on |
`TOKEN_ENDPOINT` and `INDIELOGIN_URL` are v1 settings; if present they are
ignored with a warning.
### Paths, runtime and logging
| Variable | Default | Notes |
|----------|---------|-------|
| `DATA_PATH` | `./data` (image: `/data`) | Root of all stored data |
| `DATABASE_PATH` | `./data/starpunk.db` (image: `/data/starpunk.db`) | |
| `NOTES_PATH` | `./data/notes` (image: `/data/notes`) | Created at start-up, but note files are always written to `DATA_PATH/notes` |
| `FLASK_ENV` | `production` | Reported in `/health` and admin diagnostics; the image sets `production` |
| `FLASK_DEBUG` | `0` | Debug off unless `1` (`flask run --debug` sets it locally); the image sets `0` |
| `LOG_LEVEL` | `INFO` | `DEBUG`, `INFO`, `WARNING`, `ERROR`, `CRITICAL` |
| `VERSION` | package `__version__` | Overrides the reported version; normally unset |
| `BACKGROUND_WORKERS_ENABLED` | `true` | `false` stops the webmention, WebSub and syndication workers; set in the MCP sidecar (v2.15.0) |
### Feeds
| Variable | Default | Notes |
|----------|---------|-------|
| `FEED_MAX_ITEMS` | `50` | Notes per feed |
| `FEED_CACHE_SECONDS` | `300` | Cache TTL and `Cache-Control: max-age` |
| `FEED_CACHE_ENABLED` | `true` | |
| `FEED_CACHE_MAX_SIZE` | `50` | Cached feed documents per process |
### Webmention
| Variable | Default | Notes |
|----------|---------|-------|
| `WEBMENTION_ENABLED` | `true` | Endpoint, `Link` header, showing responses on notes (not sending) |
| `WEBMENTION_WORKER_ENABLED` | `true` | Verifies received and sends outgoing webmentions; saves avatars; sends response notifications |
| `WEBMENTION_POLL_INTERVAL` | `10` | Seconds |
| `WEBMENTION_ERROR_BACKOFF` | `60` | Seconds after a worker error |
| `WEBMENTION_BATCH_SIZE` | `5` | Items per pass |
| `WEBMENTION_AUTO_APPROVE` | `true` | Valid webmentions are shown without moderation |
| `WEBMENTION_RATE_LIMIT_ENABLED` | `true` | |
| `WEBMENTION_RATE_LIMIT_IP` | `60` | Requests per window per client IP |
| `WEBMENTION_RATE_LIMIT_DOMAIN` | `10` | Requests per window per source domain |
| `WEBMENTION_RATE_LIMIT_WINDOW` | `60` | Seconds |
| `TRUSTED_PROXY` | `false` | Use `X-Forwarded-For` for the client IP (see above) |
### WebSub
| Variable | Default | Notes |
|----------|---------|-------|
| `WEBSUB_ENABLED` | `true` | Discovery links, hub and worker |
| `WEBSUB_HUB_URL` | empty | Use an external hub (StarPunk then sends publish pings) instead of the built-in one |
| `WEBSUB_MAX_SUBSCRIPTIONS` | `1000` | Total subscriptions |
| `WEBSUB_MAX_PER_HOST` | `25` | Subscriptions per callback host |
| `WEBSUB_POLL_INTERVAL` | `5` | Seconds |
### Crossposting (POSSE)
A target is enabled when its credentials are set; the syndication worker
starts only if at least one is.
| Variable | Default | Notes |
|----------|---------|-------|
| `GOTOSOCIAL_URL`, `GOTOSOCIAL_ACCESS_TOKEN` | empty | GoToSocial or any Mastodon-API server. Token scopes: `write:statuses write:media read:search write:favourites`, plus `read:statuses read:accounts` (or `read`) for backfeed |
| `GOTOSOCIAL_NAME` | `GoToSocial` | Label in the UI and `syndicate-to` |
| `GOTOSOCIAL_VISIBILITY` | `public` | |
| `GOTOSOCIAL_MAX_CHARS` | `5000` | |
| `GOTOSOCIAL_MAX_MEDIA` | `4` | |
| `BLUESKY_HANDLE`, `BLUESKY_APP_PASSWORD` | empty | |
| `BLUESKY_SERVICE` | `https://bsky.social` | |
| `BLUESKY_NAME` | `Bluesky` | |
| `BLUESKY_MAX_CHARS` | `300` | |
| `THREADS_ACCESS_TOKEN` | empty | Long-lived token; refreshed automatically into `threads_token.json`. Backfeed also needs `threads_read_replies` and `threads_manage_insights` |
| `THREADS_USERNAME` | empty | |
| `THREADS_NAME` | `Threads` | |
| `THREADS_MAX_CHARS` | `500` | |
| `THREADS_MAX_MEDIA` | `4` | |
| `SYNDICATION_POLL_INTERVAL` | `10` | Seconds |
| `BACKFEED_ENABLED` | `false` | Bring responses to crossposts back to notes (v2.13.0); needs read scopes on the GoToSocial and Threads tokens |
| `BACKFEED_MAX_AGE_DAYS` | `90` | Days after posting to keep checking a crosspost's responses |
### Notifications
| Variable | Default | Notes |
|----------|---------|-------|
| `NOTIFY_URLS` | empty (off) | Apprise URLs, space- or comma-separated (v2.14.0) |
| `NOTIFY_EVENTS` | `reply,mention,quote,problems` | Also `like`, `repost`, `bookmark` |
| `NOTIFY_PRIORITY` | `normal` | Added to URLs that don't set a priority |
### Debugging
The `METRICS_*` variables were removed in v2.12.0 and are ignored if set.
| Variable | Default | Notes |
|----------|---------|-------|
| `DEBUG_SAVE_FAILED_UPLOADS` | `false` | Keep failed uploads in `debug/` |
| `DEBUG_FILE_MAX_AGE_DAYS` | `7` | |
| `DEBUG_FILE_MAX_SIZE_MB` | `100` | |
Fixed values: the maximum request body is 50 MB (`MAX_CONTENT_LENGTH`). The
connection pool size and timeout (`DB_POOL_SIZE`, `DB_TIMEOUT`) are not read
from the environment.
## Running locally
```bash
uv venv --python 3.14 && uv pip install -r requirements-dev.txt
cp .env.example .env # set SESSION_SECRET; DEV_MODE=true and DEV_ADMIN_ME for local login
uv run flask --app app.py run --debug # or: uv run gunicorn app:app
```
`compose.yaml` builds the image from this checkout (`starpunk:local`) and runs
it with `./container-data` mounted at `/data`.
---
**Last Updated**: 2026-11-10 (v2.14.0)
+549
View File
@@ -0,0 +1,549 @@
# IndieWeb and W3C Protocols
How StarPunk implements each protocol it speaks, as of v2.14.0, with its
compliance status and the gaps that remain. Each section links to the
decision records and review documents behind it.
| Protocol | Role | Status | Main code |
|---|---|---|---|
| [Micropub](#micropub) | Server (posts and media) | Reviewed against the Recommendation in v2.9.0; deliberate deviations listed | `starpunk/micropub.py`, `starpunk/routes/micropub.py` |
| [IndieAuth](#indieauth) | Authorization server for Micropub clients | Authorization code + PKCE, metadata, revocation | `starpunk/indieauth.py`, `starpunk/routes/indieauth.py` |
| [Webmention](#webmention) | Sender and receiver | Every MUST and SHOULD checked in v2.7.0 and v2.7.2; known limits listed | `starpunk/webmention_*.py`, `starpunk/routes/webmention.py` |
| [Microformats2](#microformats2) | Publisher and consumer | Reviewed in v2.8.0 | `templates/`, `starpunk/mf2_utils.py`, `starpunk/author_discovery.py` |
| [WebSub](#websub) | Publisher and built-in hub (own topics only) | Adversarial review in v2.10.1 | `starpunk/websub.py`, `starpunk/websub_worker.py`, `starpunk/routes/websub.py` |
| [Feeds](#feeds) | RSS 2.0, Atom 1.0, JSON Feed 1.1, OPML 2.0 | Stable | `starpunk/feeds/`, `starpunk/routes/public.py` |
| Backfeed | Responses to crossposts read from the silo APIs and stored as webmentions (v2.13.0); see [syndication-architecture.md](syndication-architecture.md#backfeed-v2130-adr-073) | ADR-073 | `starpunk/backfeed.py` |
| [Web actions](#moderation-and-display) | `<indie-action>` on the like, repost and reply counts (v2.13.2) | Markup only | `templates/partials/responses.html` |
Crossposting to other networks (POSSE) is covered in
[syndication-architecture.md](syndication-architecture.md).
## Shared infrastructure: the SSRF-safe fetcher
Webmention (fetching sources, discovering endpoints, sending), WebSub
(verification and delivery), Micropub (downloading remote photos), IndieAuth
(client pages), reply contexts, crosspost link cards (v2.11.0) and response
avatars (v2.13.0; except those on StarPunk's own GoToSocial) all make
requests to URLs chosen by someone else. They use
`starpunk/webmention_ssrf.py`:
- `validate_url_for_fetch()`: http(s) only. Every address from
`getaddrinfo` must be public. Private, loopback, link-local, multicast,
reserved, unspecified, CGNAT (`100.64.0.0/10`), `0.0.0.0/8` and
IPv4-mapped forms are blocked. It returns the checked address, preferring
IPv4. `DEV_MODE` allows localhost.
- **DNS pinning** (v2.7.2): each request connects to the checked IP, with
the original host in the `Host` header and as the TLS SNI name. The
certificate is still verified against the real hostname, so a rebinding
DNS server can't send the second lookup elsewhere.
- `safe_get()` follows up to 10 redirects by hand and checks each hop. The
response is capped at 1 MB by default.
- `safe_post()` doesn't follow redirects.
- **Hard deadlines** (v2.10.1): the whole exchange, not each read, is
limited to 10 s. When the deadline passes, the client is closed, so a
server trickling bytes can't hold a worker.
- DNS resolution failures raise `ResolutionError` (a `ValueError`), which
callers can treat as temporary.
- User-Agent: `StarPunk/<version> Webmention (+<SITE_URL>)`.
Workers that drain queues (Webmention, WebSub) run as one daemon thread per
gunicorn process. Each takes a non-blocking `flock` on a lock file in the
data directory, so only one process works the queue at a time and another
takes over if it dies.
---
## Micropub
Spec: https://www.w3.org/TR/micropub/. Decisions:
[ADR-028](../decisions/ADR-028-micropub-implementation.md),
[ADR-029](../decisions/ADR-029-micropub-indieauth-integration.md),
[ADR-057](../decisions/ADR-057-media-attachment-model.md) (photos),
[ADR-067](../decisions/ADR-067-micropub-update-delete.md) (update/delete),
[ADR-068](../decisions/ADR-068-likes-and-reposts.md) (likes/reposts).
Review: [v2.9.0 Micropub review](../design/v2.9.0/2026-11-03-micropub-review.md).
### Endpoints and discovery
| Endpoint | Purpose |
|---|---|
| `GET /micropub?q=...` | Queries |
| `POST /micropub` | Create, update, delete, undelete |
| `POST /micropub/media` (and legacy `POST /media`) | Media endpoint |
Every page has `<link rel="micropub">` plus the IndieAuth discovery links
(see [IndieAuth](#indieauth)). Authentication, token placement, scopes and
error codes are described in
[simplified-auth-architecture.md §3](simplified-auth-architecture.md#3-token-use-at-the-micropub-and-media-endpoints).
### Create
Accepts form-encoded, multipart (with photo files) and JSON bodies. Only
`h=entry` / `type: ["h-entry"]` is accepted; other types get 400. Success is
**201** with `Location` set to the permalink `/note/<slug>`.
| Property / command | Handling |
|---|---|
| `content` | String, or `{"html": ...}` / `{"text": ...}`. Text is preferred. HTML-only content is stored with `content_format = 'html'` (migration 015) and rendered through `render_markdown()` (Markdown, then nh3 allowlist sanitization) |
| `summary` | Used as the text when there is no `content` |
| `name` | Stored as a `# name` heading on the first line (StarPunk has no title column) |
| `category` | Tags (strings only; person tags as h-card objects are ignored) |
| `in-reply-to`, `like-of`, `repost-of`, `bookmark-of` | URL or h-cite. At most one of like/repost/bookmark. Content is optional for these |
| `photo` | URLs, `{"value", "alt"}` objects, or uploaded files; at most 4 (ADR-057), duplicates collapsed. Own `/media/` URLs are looked up; other URLs are downloaded through `safe_get` (20 MB cap). Any photo that can't be resolved means 400 and nothing is created |
| `published` | ISO 8601; converted to naive UTC; invalid gets 400 |
| `post-status` | `published` (default) or `draft`; others get 400 |
| `visibility` | `public`; `private` and `unlisted` are saved as drafts (StarPunk has no private posts); others get 400 |
| `video`, `audio` | 400 (not supported) |
| `location`, `checkin`, others | Ignored (allowed by 3.3.5) |
| `mp-slug` | Custom slug |
| `mp-syndicate-to` | Crosspost targets. **Absent means none** (v2.9.0). See [syndication-architecture.md](syndication-architecture.md#choosing-targets) |
`mp-*` keys are read as commands from the top level or from inside JSON
`properties`, and are never stored. JSON property values must be arrays; a
bare scalar is wrapped on create. If anything fails after the note is
created, the note is removed and 500 is returned, so the client never
believes a half-made post failed.
Post types advertised in `q=config`: note, article, photo, reply, like,
repost, bookmark.
### Update, delete, undelete
- **Update** (JSON only, `update` scope) targets a `/note/<slug>` URL on
this site.
- `replace` works on `content`, `name`, `category`, `in-reply-to`,
`post-status` and `photo`.
- `add` works on `category` and `photo` only.
- `delete` takes a list of property names, or values to remove for
`category` and `photo`.
- Values must be arrays. An update without any operation gets 400.
Unsupported properties get 400.
- Photos are resolved before anything changes. Success is **204**.
- Edits follow crossposts and are re-announced to Webmention targets and
WebSub subscribers.
- **Delete** (`delete` scope) is soft: the note is moved to the trash, its
page returns 410, and `undelete` restores it. Both return **204**.
### Queries
| Query | Scope | Result |
|---|---|---|
| `q=config` | any token | `media-endpoint`, `syndicate-to`, `q` list, `post-types` |
| `q=syndicate-to` | any token | `{"syndicate-to": [{"uid", "name"}]}` |
| `q=source&url=...[&properties[]=...]` | create/update/delete/draft | One post as mf2 JSON. Content comes back as `{"html": ...}` if it was authored as HTML. Includes `syndication` URLs and `post-status` |
| `q=source` without `url` | same | `{"items": [...]}`, drafts included, `limit` (default 10, max 50) and `offset` |
| `q=category[&filter=...]` | same | Tags on notes that aren't deleted (drafts count), most used first (v2.12.1) |
Unknown or missing `q` gets 400.
### Media endpoint
`POST /micropub/media`, multipart with one part named `file`, `media` or
`create` scope. It returns **201** with `Location: <SITE_URL>/media/<path>`.
Images are JPEG, PNG, GIF, WebP or HEIC, up to 50 MB in. They are optimized
and get responsive variants (`starpunk/media.py`,
[ADR-058](../decisions/ADR-058-image-optimization-strategy.md)). Validation
failures get 400.
### Known deviations and gaps
- **403 for `insufficient_scope`** (RFC 6750) where the Micropub error table
says 401. Kept on purpose.
- `add` is only supported for `category` and `photo`. `published`,
`syndication` and `location` can't be updated. `mp-syndicate-to` is only
read on create.
- `location` / `checkin` are ignored, and person tags are dropped.
- `video` and `audio` are refused.
- The `access_token` query parameter is still accepted on GET.
- Only notes (`/note/<slug>`) can be the target of update/delete/source.
---
## IndieAuth
StarPunk is its own IndieAuth authorization server for Micropub clients,
issuing tokens for the single identity `ADMIN_ME`. The admin approves
clients on a consent screen protected by OIDC login. Supported: the
authorization code flow with mandatory PKCE S256, `iss` in responses (RFC
9207), the metadata endpoint, a profile-only flow, a token endpoint, RFC
7009 revocation, and legacy token verification (`GET /auth/token`). Client
`redirect_uri`s must be same-origin or published by the client (JSON
metadata or `rel=redirect_uri`).
Full description, scopes and gaps:
[simplified-auth-architecture.md](simplified-auth-architecture.md).
Decision: [ADR-064](../decisions/ADR-064-oidc-admin-login-and-builtin-indieauth-server.md).
Known gaps: `scopes_supported` doesn't list `draft`. There are no
introspection, userinfo or refresh tokens.
---
## Webmention
Spec: https://www.w3.org/TR/webmention/ (Recommendation, 12 January 2017).
Decision: [ADR-063](../decisions/ADR-063-webmention-implementation.md).
Reviews: [v2.6.9 fixes](../design/v2.6.9/2026-11-03-webmention-processing-fixes.md),
[v2.7.0 compliance](../design/v2.7.0/2026-11-03-webmention-spec-compliance.md),
[v2.7.2 remaining SHOULDs](../design/v2.7.2/2026-11-03-webmention-shoulds.md).
### Endpoint discovery (advertising)
With `WEBMENTION_ENABLED` (default true), every HTML response carries
`Link: </webmention>; rel="webmention"`, and `base.html` has
`<link rel="webmention" href="/webmention">`. When it is off, the receiving
route isn't registered at all.
### Receiving
`POST /webmention` (`routes/webmention.py`):
1. Rate limits: 60 requests per minute per client IP and 10 per minute per
source domain (`WEBMENTION_RATE_LIMIT_*`). Over the limit gets 429 with
`Retry-After`. Counts are shared by all server processes (`rate_limits`
table, v2.11.0). The client IP
comes from `X-Forwarded-For` only when `TRUSTED_PROXY=true`.
2. Form-encoded (JSON also accepted). `source` and `target` are required,
http(s), at most 2048 characters, and different from each other.
Otherwise 400.
3. The target's host and port must be this site's. The fragment is stripped
before storing. The target must be a **published note**: anything else,
including the home page, gets 400 `invalid_target`.
4. The row is upserted into `webmentions` (unique on source + target) as
`pending`, and **202** is returned. A repeat from the same source resets
it to `pending` (an update, 3.2.4).
Verification is asynchronous (`WebmentionWorker`, poll
`WEBMENTION_POLL_INTERVAL` 10 s, batch of 5):
- **Fetch** the source with `safe_get` (redirects followed and checked,
1 MB, 10 s), with an `Accept` header preferring HTML.
- **Exact-match verification** (3.2.2), per media type:
- HTML: every link-bearing attribute (`lxml iterlinks`: `a`/`link` href,
`img`/`video`/`audio` src, ...), resolved against the final URL.
- JSON: any string value.
- Other `text/*`: substring match.
- Any other type: no match.
- Fragments are ignored on both sides.
- **Parse mf2** (HTML only) with mf2py:
- `mf2_utils.responding_entry()` picks the h-entry that responds to the
target.
- `response_type()` classifies it as reply, like, repost, bookmark or
mention.
- `entry_author()` follows the authorship algorithm without extra
fetches: author property, then the parent h-feed's author, then
`rel=author`, then the representative h-card.
- Also stored: a content snippet (e-content or p-summary, sanitized with
nh3, at most 500 characters), the source's `u-url` and `dt-published`
(migration 014).
- The author URL and photo are kept only if they are absolute http(s)
URLs (XSS, 4.1).
- **Outcomes**:
- Valid: `validation_status = 'valid'`. A first-time mention gets
`moderation_status = 'approved'` if `WEBMENTION_AUTO_APPROVE` (default
true), otherwise `pending`. An update keeps the admin's earlier
moderation decision, including `hidden` and `deleted`.
- Source returns 4xx (including 410), or no longer links: `invalid`,
which hides it (3.2.4 deletion).
- Network error, timeout or 5xx: a mention that was valid before stays
valid with its old data. A new one becomes `error`.
- A source that is one of the note's own syndicated copies
(`syndications.remote_url`) is `invalid` (v2.8.0).
- A source that backfeed has already stored for the note is `invalid`
("Already backfed from the silo", v2.13.0).
- **Re-send on display change** (3.1.4, v2.7.2): if what the note shows
for this webmention changed, the note's own outgoing webmentions are
re-queued. That covers a new response, a changed author or snippet, or a
change of shown/hidden. Unchanged re-validations send nothing, so two
sites doing the same don't ping-pong.
### Moderation and display
`/admin/webmentions` lists, filters (by status, including Deleted, and by
origin), and approves, hides or deletes responses, one at a time or in bulk;
it also has the Backfill button. Deleting keeps the row as a `deleted`
tombstone, so a re-sent webmention or backfeed doesn't bring it back.
Changes that alter what is shown also re-send the note's webmentions.
Note pages show approved, valid responses. Under the note, a line of counts
(♥ likes, 🔁 reposts, 💬 replies, ❝ quotes (v2.14.2), bookmarks) opens a card on hover or focus
listing who responded and where from, including anonymous silo counts. Each
like, repost and reply count holds an `<indie-action do with>` web action
whose fallback links to the note's crossposts (v2.13.2). The Responses
section below holds replies (`u-comment h-cite`), quotes and mentions;
backfed ones say "via <silo>". See [Microformats2](#microformats2).
### Notifications (v2.14.0, ADR-074)
With `NOTIFY_URLS` set, new valid responses (received and backfed) are sent
to the owner once through Apprise by a sweep in the webmention worker, which
sets `webmentions.notified_at` (migration 021). Likes, reposts and bookmarks
are grouped per note; which kinds are sent is set by `NOTIFY_EVENTS`
(default replies, mentions, quotes and problems). Items still unsent after
24 hours are dropped. Responses found by a backfill, or by a copy's first
check on a note older than a day, are never notified.
### Sending
Queued by `webmention_sender.queue_webmentions_for_note()` from `notes.py`.
Sending isn't gated by `WEBMENTION_ENABLED`; it runs whenever the worker
does (`WEBMENTION_WORKER_ENABLED`).
| Event | Targets |
|---|---|
| Publish | Reply / like / repost / bookmark target first, then `<a href>` links from the rendered HTML in document order (not inside `<pre>`/`<code>`), absolute http(s) only, de-duplicated, at most 10 |
| Edit (published) | The same, plus every target notified before and links the edit removed (not capped) (3.1.4) |
| Delete | Every previously notified target is re-queued (3.1.5); for a hard delete the queue rows are detached first so the cascade doesn't remove them |
| Undelete | Re-queued again |
| Displayed responses change | Re-queued (see above) |
Re-queuing resets the cached endpoint, so discovery runs again (3.1.4).
**Endpoint discovery** (3.1.2): GET the target with `safe_get` (redirects
checked). The first `Link` header with a `rel` token `webmention` wins. The
header is parsed per RFC 8288: several links per header, several headers,
rel token lists, case-insensitive. Otherwise the first `<link>` or `<a>`
with `rel~=webmention` and an `href`, in document order. Relative URLs
resolve against the final URL, an empty `href` means the target itself, and
the query string is kept.
**Send** (3.1.3): a form-encoded POST through `safe_post`, pinned and with
no redirects followed (a 301/302 would turn the POST into a GET).
| Result | Outcome |
|---|---|
| Any 2xx | `sent` |
| 429, 5xx, network error | Retried after 1 h, 6 h, 24 h; `failed` after the 4th attempt |
| Other 3xx/4xx, SSRF-blocked endpoint | `failed` |
| No endpoint | `no_endpoint` |
**Deleted posts**: `/note/<slug>` returns **410 Gone** with a tombstone page
for soft-deleted notes, and for hard-deleted notes that had sent
webmentions (3.1.5).
### Security (section 4)
- XSS: author URL and photo must be http(s). Snippets are sanitized with
nh3 (and re-sanitized after truncation). The mentions list uses
`striptags`.
- SSRF: all fetches and POSTs go through the pinned fetcher described
above, so the endpoint URL is checked before POSTing.
- DoS: rate limits on receiving, a 1 MB / 10 s cap on fetches, and
asynchronous processing.
### Known limits
From the v2.7.0 review, still current:
- Home-page mentions (and mentions of anything that isn't a published note)
are refused rather than stored.
- Discovery ignores HTTP cache headers (5.3). Targets are fetched only when
a note is created, edited or deleted.
- The 410 page is a generic tombstone without microformats.
- `error` rows (transient failures of a new mention, or unexpected errors)
are not retried automatically. They are re-checked only when the sender
re-sends.
- The 202 response has no `Location` status URL.
---
## Microformats2
Specs: microformats2 parsing, h-entry, h-card, h-feed, h-cite,
representative h-card, authorship, Post Type Discovery. Decisions:
[ADR-040](../decisions/ADR-040-microformats2-compliance.md),
[ADR-061](../decisions/ADR-061-author-discovery.md). Review:
[v2.8.0 microformats2 review](../design/v2.8.0/2026-11-03-microformats2-review.md).
### What pages publish
| Page | Markup |
|---|---|
| `/` (home) | `h-feed` with `p-name`, hidden `u-url`, hidden `p-author h-card` (photo, name, URL, note), feed `u-photo`; each note an `h-entry` preview with `u-uid` |
| `/note/<slug>` | `h-entry`: `p-name` only when the note starts with a `# Heading` (so plain notes have no implied name); `u-photo` as direct children above `e-content`; `u-url u-uid` with `dt-published` (RFC 3339 UTC `Z`); `dt-updated`; `p-category` with `rel=tag`; `u-syndication` links; hidden `p-author h-card` |
| Replies | `u-in-reply-to h-cite` holding the cached reply context (`p-content`, `u-url`) |
| Likes / reposts / bookmarks | `u-like-of` / `u-repost-of` / `u-bookmark-of` link with "★ Liked" / "🔁 Reposted" / "🔖 Bookmarked" |
| Received responses | `u-comment h-cite` (replies: `p-author h-card` with `u-url` on an inner link, `u-photo`, `p-name`; `e-content`; `u-url`; `dt-published`), likes, reposts and bookmarks as `u-like`, `u-repost`, `u-bookmark` h-cites in the counts line card; mentions and quotes as `u-mention` h-cites |
| `/likes`, `/tag/<tag>` | `h-feed` with `p-name`, `u-url`, `p-author h-card` |
| Response counts | `/`, `/tag/<tag>` and `/likes` show each note's like, repost and reply counts, linking to its `#responses` (v2.13.4) |
Post Type Discovery gives note, article, reply, like, repost, bookmark and
photo correctly.
### Identity links
`base.html` always emits `rel=me` and `rel=author` pointing at the author's
home page (`author.url`, else `author.me`, else `ADMIN_ME`), even on a fresh
install (v2.8.0). Further `rel=me` links come from the discovered profile.
Because the `p-author h-card`'s `u-url` matches the `rel=me` URL, parsers
treat it as the page's representative h-card. Without a discovered profile
there is no h-card, only the `rel` links.
### Consuming
- **Author profile** (`author_discovery.py`, ADR-061): on login, StarPunk
fetches `ADMIN_ME`, picks its representative h-card with
`mf2_utils.representative_hcard()` (uid and url match the page, then
url matches a `rel=me`, then a single h-card whose url is the page),
parsed with the URL after redirects. It stores name, photo, URL, note and
rel-me links, cached for 24 hours. If discovery fails, the stale cache is
used.
- **Webmention sources**: `responding_entry`, `response_type` and
`entry_author` (see [Webmention](#webmention)).
- **Reply context** (`reply_context.py`): title, snippet, author and site
name of the post being replied to, liked, reposted or bookmarked, cached
in `reply_contexts`.
### Known gaps
From the v2.8.0 review, "Not done":
- The reply-context `h-cite` doesn't carry the cached author h-card, name or
published time.
- Photo alt text falls back to "Image" when a photo-only post has no
caption.
- Person tags (`p-category h-card`) aren't supported.
- Webmentions received before v2.8.0 keep NULL source details until they
are re-validated.
---
## WebSub
Spec: https://www.w3.org/TR/websub/. Decision:
[ADR-071](../decisions/ADR-071-websub.md), including its v2.10.1
amendments, which replace the original text where they differ.
### Topics
All absolute URLs under `SITE_URL`:
| Topic | Content type | Updated when |
|---|---|---|
| `/` | text/html | A public non-like note changes |
| `/feed.atom`, `/feed.rss`, `/feed.json` | feed types | A public non-like note changes |
| `/likes` | text/html | A public like changes |
| `/tag/<name>` | text/html | A note with that tag (before or after the change) changes |
`/feed.xml` (legacy) advertises `/feed.rss` as its `rel=self`. The
negotiated `/feed` advertises the topic of the format it served.
### Discovery
Topic responses (200/304) carry `Link: <hub>; rel="hub"` and
`Link: <topic>; rel="self"`. The same links are embedded: `<link rel="hub">`
and `<link rel="self">` in HTML heads, a hub `<link>` in Atom, an
`<atom:link rel="hub">` in RSS, and `"hubs"` in JSON Feed. All of this
happens only while `WEBSUB_ENABLED` (default true).
### On change
`websub.note_changed()` runs on create, edit, publish/unpublish, delete and
undelete, and once photos are attached. If the note is or was public, it:
1. bumps `content_version`, which is in the feed cache key and the ETag, so
every process serves fresh feeds at once;
2. queues one delivery per active subscription for each affected topic.
Only the latest content is sent: a newer delivery replaces a waiting one.
Drafts that stay drafts change nothing.
### Hub
`POST /websub`, form-encoded only (case-insensitive check):
- `hub.mode` is `subscribe` or `unsubscribe`.
- `hub.topic` must be one of the topics above, after decoding unreserved
percent-encodings and comparing scheme and host case-insensitively. Tag
topics must name an existing tag.
- `hub.callback` must:
- be printable, at most 2,048 characters, and absolute http(s);
- not be on this site;
- use port 80, 443 or 1024 and up;
- pass the SSRF check.
It is stored with an IDNA host.
- `hub.secret` must be non-empty and under 200 bytes. `hub.lease_seconds`
is clamped to 1 to 30 days (default 10). Both are ignored for
unsubscriptions.
- The hub answers **202**, then verifies later. Errors are 4xx with a
plain-text reason.
Limits (429 when hit):
- 30 requests per minute per client IP (shared by all processes; needs
`TRUSTED_PROXY=true` behind a proxy);
- at most 200 pending verifications;
- `WEBSUB_MAX_SUBSCRIPTIONS` (1000) in total;
- `WEBSUB_MAX_PER_HOST` (25) per callback IP address.
The limits count pending requests too. Renewals are always allowed.
`WebSubWorker` (file lock, poll `WEBSUB_POLL_INTERVAL` 5 s) handles each
queued item on its own, so one bad item can't stop it:
- **Verification of intent** (5.3): `GET` the callback (existing query
kept) with `hub.mode`, the topic exactly as requested, `hub.challenge`
and `hub.lease_seconds`. Success is a 2xx whose body equals the challenge.
Replies over 4 KB and redirects fail. Network/DNS errors and 5xx are
retried after 1, 5 and 15 minutes. A request id per intent (migration
017) stops an older verification from consuming a newer request. A
subscription row exists only once verified.
- **Distribution** (7): the topic is rendered inside the app, once per pass,
and POSTed with its `Content-Type`, `Link` rel=hub/self, and
`X-Hub-Signature: sha256=<HMAC>` when there is a secret.
- 2xx: success.
- 410: removes the subscription.
- Other failures: retried after 1 min, 5 min, 30 min, 2 h and 6 h, then
that delivery is dropped. The subscription stays until its lease ends.
- A topic that renders 404/410 ends its subscriptions with a
`hub.mode=denied` notice.
- Expired subscriptions are pruned hourly.
Outbound requests use `safe_get` / `safe_post`: pinned, no redirects,
10 s hard deadline.
**External hub**: with `WEBSUB_HUB_URL` set, StarPunk advertises that hub
and sends it a `hub.mode=publish` ping per changed topic. Its own `/websub`
refuses new subscriptions but still accepts unsubscriptions.
The admin dashboard shows subscribers per topic and deliveries being
retried.
### Known limits
- The hub serves only this site's topics, by design. websub.rocks' hub
tests use their own topics, so they can't be run against it.
- Deliveries always carry the full topic content (no diffs). Fat pings
aren't sent to external hubs.
---
## Feeds
Decisions: [ADR-014](../decisions/ADR-014-rss-feed-implementation.md),
[ADR-038](../decisions/ADR-038-syndication-formats.md),
[ADR-059](../decisions/ADR-059-full-feed-media-standardization.md).
| URL | Format |
|---|---|
| `/feed.rss` | RSS 2.0 (with `atom:link rel=self` and Media RSS) |
| `/feed.xml` | Legacy alias of `/feed.rss` |
| `/feed.atom` | Atom 1.0 |
| `/feed.json` | JSON Feed 1.1 |
| `/feed` | Content negotiation on `Accept` (RSS default; 406 with `X-Available-Formats` if nothing fits; `Vary: Accept`) |
| `/opml.xml` | OPML 2.0 listing the three feeds |
- Feeds contain the newest `FEED_MAX_ITEMS` (50) published notes, with full
HTML content and media. Likes are excluded, as on the home page.
Reposts and bookmarks start with a "🔁 Reposted" / "🔖 Bookmarked" line.
- Feeds are cached for `FEED_CACHE_SECONDS` (300). The cache key and ETag
include `content_version`, so any change to a public note invalidates
them in every process. `If-None-Match` gets 304.
- `base.html` advertises the negotiated `/feed` (as RSS) and the OPML file
with `rel=alternate`. Atom and JSON Feed are listed in the OPML but have
no `<link rel=alternate>` of their own.
- WebSub hub links are embedded in each feed (see above).
Gaps: there are no per-tag feeds (tag pages are h-feeds and WebSub topics,
but have no RSS/Atom/JSON equivalent). The `/likes` page has no feed.
+77
View File
@@ -0,0 +1,77 @@
# MCP server (v2.15.0)
How AI clients use StarPunk through the
[Model Context Protocol](https://modelcontextprotocol.io/). The decision and
alternatives are in [ADR-075](../decisions/ADR-075-mcp-server.md).
## Shape
```
client ──HTTPS──> reverse proxy
/mcp*, /.well-known/oauth-protected-resource* ──> starpunk-mcp:8001 (uvicorn, MCP SDK)
everything else (incl. /auth/*, /.well-known/oauth-authorization-server)
──> starpunk:8000 (gunicorn, Flask)
starpunk and starpunk-mcp: same image, same environment, same /data volume
```
| Module | Role |
|---|---|
| `starpunk/mcp_server/app.py` | ASGI entry: `uvicorn starpunk.mcp_server.app:app --host 0.0.0.0 --port 8001` |
| `starpunk/mcp_server/server.py` | `create_mcp_server(app)` and `create_asgi_app(app=None)`: stateless Streamable HTTP, JSON responses, 16 MB request limit, DNS rebinding protection |
| `starpunk/mcp_server/auth.py` | `StarPunkTokenVerifier`: StarPunk tokens to SDK `AccessToken`s |
| `starpunk/mcp_server/tools.py` | The tools; each runs in a thread inside `app.test_request_context(base_url=SITE_URL)` |
`create_asgi_app()` with no argument builds the Flask app from the
environment with `BACKGROUND_WORKERS_ENABLED=false`. The sidecar container
should also set that variable, for clarity.
## Authorization flow
1. The client POSTs to `/mcp` without a token and gets `401` with
`WWW-Authenticate: Bearer ... resource_metadata="SITE_URL/.well-known/oauth-protected-resource/mcp"`.
2. That document names `SITE_URL/mcp` as the resource and `SITE_URL/` as
the authorization server. The client reads
`/.well-known/oauth-authorization-server`, which says it accepts client ID
metadata documents and public clients.
3. The client sends the owner to `/auth/authorize` with its `client_id` URL,
PKCE and `resource=SITE_URL/mcp`. StarPunk fetches the client's metadata
document to check `redirect_uri` (unless it has the same origin as the
`client_id`), and the owner logs in with OIDC and approves. With no
`scope`, the default MCP scopes are offered.
4. The token endpoint exchanges the code. The token is bound to the
resource.
5. Each MCP request carries the token. The verifier looks it up and the SDK
refuses it unless its resource is `SITE_URL/mcp`.
Details of codes, tokens and scopes are in
[simplified-auth-architecture.md](simplified-auth-architecture.md).
## Tools
| Tool | Does | Scope |
|---|---|---|
| `search_notes(query, include_drafts, limit)` | FTS5 search, with snippets | any; drafts need a posting scope |
| `list_notes(include_drafts, limit, offset)` | Newest first, summaries | any; drafts need a posting scope |
| `get_note(url_or_slug)` | Microformats2 properties (as Micropub `q=source`) | any; drafts need a posting scope |
| `list_tags(filter)` | Tags in use, most used first | posting scope |
| `get_site_config()` | Micropub `q=config`, including `syndicate-to` targets | any |
| `create_note(content, title, tags, draft=True, photos, in_reply_to, like_of, repost_of, bookmark_of, syndicate_to, slug, published_at)` | Micropub create | `create` or `draft` |
| `update_note(url, content, title, tags, add_tags, remove_tags, draft, photos, in_reply_to)` | Micropub update | `update` |
| `delete_note(url)`, `undelete_note(url)` | Soft delete and restore | `delete` |
| `upload_media(filename, data_base64)` | Save an image, return its URL | `media` or `create` |
"Posting scope" means `create`, `update`, `delete` or `draft`. Limits are
capped at 100. Micropub errors come back as tool errors (`isError`) with the
Micropub error description.
## Operating it
- Environment: the same as the main container, plus
`BACKGROUND_WORKERS_ENABLED=false`.
- Proxy: route `/mcp*` and `/.well-known/oauth-protected-resource*` to port
8001, keeping the `Host` header (DNS rebinding protection checks it
against `SITE_URL`).
- Connecting: claude.ai Settings → Connectors → Add custom connector with
`SITE_URL/mcp`; Claude Code: `claude mcp add --transport http starpunk
SITE_URL/mcp`, then `/mcp` to authenticate.
- Revoking: Admin → Connected Apps (MCP tokens are marked "(MCP)").
File diff suppressed because it is too large. Load diff
+308 -380
View File
@@ -1,428 +1,356 @@
# StarPunk Simplified Authentication Architecture
# Authentication and Authorization Architecture
Current as of v2.14.0 (no auth changes since v2.11.0). Decision record: [ADR-064](../decisions/ADR-064-oidc-admin-login-and-builtin-indieauth-server.md),
which supersedes the "no self-hosted IndieAuth" decisions
([ADR-050](../decisions/ADR-050-remove-custom-indieauth-server.md),
[ADR-056](../decisions/ADR-056-no-selfhosted-indieauth.md)) and the external
token verification of ADR-030 / ADR-036.
> The filename is historical: it described the v1.x "external provider only"
> design. Other documents link to it, so it is kept.
## Overview
After removing the custom IndieAuth authorization server, StarPunk becomes a pure Micropub server that relies on external providers for all authentication and authorization.
StarPunk is a single-user site with two kinds of client and two auth
mechanisms:
## Architecture Diagrams
| Who | Mechanism | Credential | Code |
|---|---|---|---|
| The site owner in a browser (admin UI, consent screen) | OpenID Connect login against an external provider (e.g. Pocket ID) | `starpunk_session` cookie, backed by a hashed row in `sessions` | `starpunk/auth.py`, `starpunk/routes/auth.py` |
| Micropub apps (Quill, Indigenous, ...) | StarPunk's own IndieAuth authorization server | Bearer access token, backed by a hashed row in `indieauth_tokens` | `starpunk/indieauth.py`, `starpunk/routes/indieauth.py` |
The two are chained: a Micropub app is approved on a consent screen that
requires an admin session, so every token ultimately rests on the OIDC
login. StarPunk stores no passwords.
### Before: Complex Mixed-Mode Architecture
```
┌──────────────────────────────────────────────────────────────┐
│ StarPunk Instance │
│ │
│ ┌────────────────────────────────────────────────────────┐ │
│ │ Web Interface │ │
│ │ ┌─────────────┐ ┌──────────────┐ ┌──────────────┐ │ │
│ │ │ Admin Login │ │ Authorization │ │ Token Issuer │ │ │
│ │ └─────────────┘ └──────────────┘ └──────────────┘ │ │
│ └────────────────────────────────────────────────────────┘ │
│ │
│ ┌────────────────────────────────────────────────────────┐ │
│ │ Auth Module │ │
│ │ ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐ │ │
│ │ │ Sessions │ │ PKCE │ │ Tokens │ │ Codes │ │ │
│ │ └──────────┘ └──────────┘ └──────────┘ └──────────┘ │ │
│ └────────────────────────────────────────────────────────┘ │
│ │
│ ┌────────────────────────────────────────────────────────┐ │
│ │ Database │ │
│ │ ┌────────┐ ┌──────────────────┐ ┌─────────────────┐ │ │
│ │ │ Users │ │ authorization_codes│ │ tokens │ │ │
│ │ └────────┘ └──────────────────┘ └─────────────────┘ │ │
│ └────────────────────────────────────────────────────────┘ │
└──────────────────────────────────────────────────────────────┘
Browser (owner) StarPunk OIDC provider
─────────────── ──────── ─────────────
POST /auth/login ─────────────▶ initiate_login() ── redirect ──▶ authorize (PKCE, nonce)
owner authenticates
GET /auth/callback?code&state ◀────────────────────── redirect ───┘
handle_callback(): code exchange, ID token checks,
sub == OIDC_ADMIN_SUB → create_session(ADMIN_ME)
◀── Set-Cookie: starpunk_session
Problems:
- 500+ lines of security-critical code
- Dual role: authorization server AND resource server
- Complex token lifecycle management
- Database bloat with token storage
- Maintenance burden for security updates
Micropub app StarPunk
──────────── ────────
discovers indieauth-metadata / authorization_endpoint / token_endpoint
GET /auth/authorize?... ────────▶ validate request → (OIDC login if no session) → consent page
owner approves ─────────────────▶ POST /auth/authorize/approve → redirect: code, state, iss
POST /auth/token (code, verifier) ▶ access token (hash stored in indieauth_tokens)
POST /micropub, /micropub/media ─▶ verify_access_token() + scope check
```
### After: Clean Separation of Concerns
```
┌──────────────────────────────────────────────────────────────┐
│ StarPunk Instance │
│ │
│ ┌────────────────────────────────────────────────────────┐ │
│ │ Web Interface │ │
│ │ ┌─────────────┐ ┌──────────────┐ │ │
│ │ │ Admin Login │ │ Micropub │ │ │
│ │ └─────────────┘ └──────────────┘ │ │
│ └────────────────────────────────────────────────────────┘ │
│ │
│ ┌────────────────────────────────────────────────────────┐ │
│ │ Auth Module │ │
│ │ ┌──────────────┐ ┌──────────────────────┐ │ │
│ │ │ Sessions │ │ Token Verification │ │ │
│ │ │ (Admin Only) │ │ (External Provider) │ │ │
│ │ └──────────────┘ └──────────────────────┘ │ │
│ └────────────────────────────────────────────────────────┘ │
│ │
│ ┌────────────────────────────────────────────────────────┐ │
│ │ Database │ │
│ │ ┌────────┐ ┌──────────┐ ┌─────────┐ │ │
│ │ │ Users │ │auth_state│ │ posts │ (No token tables)│ │
│ │ └────────┘ └──────────┘ └─────────┘ │ │
│ └────────────────────────────────────────────────────────┘ │
└──────────────────────────────────────────────────────────────┘
│
│ API Calls
▼
┌──────────────────────────────────────────────────────────────┐
│ External IndieAuth Providers │
│ ┌─────────────────────┐ ┌─────────────────────────┐ │
│ │ indieauth.com │ │ tokens.indieauth.com │ │
│ │ (Authorization) │ │ (Token Verification) │ │
│ └─────────────────────┘ └─────────────────────────┘ │
└──────────────────────────────────────────────────────────────┘
## 1. Admin login (OIDC)
Benefits:
- 500+ lines of code removed
- Clear single responsibility
- No security burden
- Minimal database footprint
- Zero maintenance for auth code
```
### Flow
## Authentication Flows
1. `GET /auth/login` shows `templates/admin/login.html`, or redirects to the
dashboard if the session cookie is already valid.
2. `POST /auth/login` calls `auth.initiate_login(next)`:
- Fetches `{OIDC_ISSUER}/.well-known/openid-configuration` (cached for an
hour per process). The document's `issuer` must equal `OIDC_ISSUER`, and
it must name `authorization_endpoint`, `token_endpoint` and `jwks_uri`.
- Generates `state`, `nonce` and a PKCE `code_verifier`, and stores them in
`oidc_login_state` with the sanitized `next` URL (same site only) and a
10-minute expiry.
- Redirects to the provider with `response_type=code`, `scope=openid` and
`code_challenge_method=S256`.
3. `GET /auth/callback`:
- A provider `error` parameter, or a missing `code`/`state`, flashes an
error and returns to the login page.
- The `state` row is fetched and deleted in one step: it is single use and
must not have expired.
- If the provider sent `iss` (RFC 9207), it must match the issuer.
- The code is exchanged at the token endpoint along with the
`code_verifier`. StarPunk authenticates as a confidential client (HTTP
Basic, `OIDC_CLIENT_ID` / `OIDC_CLIENT_SECRET`).
- The ID token is verified with PyJWT against the provider's JWKS. The
checks are: the signature (algorithms from the provider metadata, never
`none`), `iss`, `aud` equal to the client id, `exp`, required `iat` and
`sub`, a 60 s leeway, and a `nonce` equal to the stored one.
- `sub` must equal `OIDC_ADMIN_SUB` (constant-time comparison). Any other
account gets "Not authorized as admin".
- A session is created for `ADMIN_ME`, the owner's identity URL. It is also
the `me` of every issued token. The author h-card is refreshed from
`ADMIN_ME` on a best-effort basis that never blocks login (see
[ADR-061](../decisions/ADR-061-author-discovery.md)).
### Flow 1: Admin Authentication (Unchanged)
```
Admin User StarPunk IndieLogin.com
│ │ │
├──── GET /admin/login ───→ │ │
│ │ │
│ ←── Login Form ─────────── │ │
│ │ │
├──── POST /auth/login ───→ │ │
│ (me=admin.com) │ │
│ ├──── Redirect ──────────────→ │
│ │ (client_id=starpunk.com) │
│ ←──────────── Authorization Request ───────────────────── │
│ │ │
├───────────── Authenticate with IndieLogin ──────────────→ │
│ │ │
│ │ ←── Callback ────────────────│
│ │ (me=admin.com) │
│ │ │
│ ←── Session Cookie ─────── │ │
│ │ │
│ Admin Access │ │
```
### Sessions
### Flow 2: Micropub Client Authentication (Simplified)
```
Micropub Client StarPunk External Token Endpoint
│ │ │
├─── POST /micropub ───→ │ │
│ Bearer: token123 │ │
│ ├──── GET /token ─────────→ │
│ │ Bearer: token123 │
│ │ │
│ │ ←── Token Info ──────────│
│ │ {me, scope, client_id} │
│ │ │
│ │ [Validate me==ADMIN_ME] │
│ │ [Check scope includes │
│ │ "create"] │
│ │ │
│ ←── 201 Created ────────│ │
│ Location: /post/123 │ │
```
- `create_session(me)` makes a random 32-byte URL-safe token. Only its
SHA-256 hash is stored in `sessions`, with `me`, `expires_at`
(`SESSION_LIFETIME` days, default 30), the user agent and the IP. Expired
sessions and login states are cleaned up each time a session is created.
- Cookie `starpunk_session`: `HttpOnly`, `SameSite=Lax`, `Secure` when
`SITE_URL` is https, `max-age` equal to the session lifetime.
- `verify_session(token)` looks up the hash, checks expiry and updates
`last_used_at`.
- `@require_auth` protects `/admin/*`. Without a valid session it remembers
the URL and redirects to `/auth/login`. It sets `g.user` and `g.me`.
- A `before_request` hook also resolves the session on public pages that
carry the cookie, so the header can show "Admin" instead of "Log in".
- `POST /auth/logout` deletes the session row and the cookie.
## Component Responsibilities
### Configuration
### StarPunk Components
Outside `DEV_MODE`, `OIDC_ISSUER`, `OIDC_CLIENT_ID`, `OIDC_CLIENT_SECRET`,
`OIDC_ADMIN_SUB` and `ADMIN_ME` are required, and startup fails without them.
The provider must have `{SITE_URL}/auth/callback` registered as a redirect
URI.
#### 1. Admin Authentication (`/auth/*`)
**Responsibility**: Manage admin sessions via IndieLogin.com
**Does**:
- Initiate OAuth flow with IndieLogin.com
- Validate callback and create session
- Manage session lifecycle
### Development mode
**Does NOT**:
- Issue tokens
- Store passwords
- Manage user identities
With `DEV_MODE=true` (and `DEV_ADMIN_ME`), `/dev/login` creates a session
without OIDC. The blueprint is registered only in dev mode, and its
`before_request` guard returns 404 if `DEV_MODE` is off. Every page shows a
warning banner while it is on. Never enable it in production.
#### 2. Micropub Endpoint (`/micropub`)
**Responsibility**: Accept and process Micropub requests
**Does**:
- Extract Bearer tokens from requests
- Verify tokens with external endpoint
- Create/update/delete posts
- Return proper Micropub responses
## 2. Built-in IndieAuth authorization server
**Does NOT**:
- Issue tokens
- Manage authorization codes
- Store token data
StarPunk issues tokens for exactly one identity, `ADMIN_ME`. It implements
the authorization code flow of the current IndieAuth spec
(https://indieauth.spec.indieweb.org/), with PKCE required.
#### 3. Token Verification Module
**Responsibility**: Validate tokens with external providers
**Does**:
- Call external token endpoint
- Cache valid tokens (5 min TTL)
- Validate scope and identity
### Endpoints
**Does NOT**:
- Generate tokens
- Store tokens permanently
- Manage token lifecycle
| Method | Path | Purpose |
|---|---|---|
| GET | `/.well-known/oauth-authorization-server` | Server metadata (the target of `rel=indieauth-metadata`) |
| GET | `/auth/authorize` | Validate the request; show the consent screen (admin session required) |
| POST | `/auth/authorize/approve` | Consent decision from the screen |
| POST | `/auth/authorize` | Profile-only code redemption, returns `{"me": ...}` |
| POST | `/auth/token` | Exchange a code for an access token; also legacy `action=revoke` |
| GET | `/auth/token` | Legacy token verification (Bearer token → `me`, `client_id`, `scope`) |
| POST | `/auth/revoke` | RFC 7009 revocation (always 200) |
### External Provider Responsibilities
All JSON responses carry `Cache-Control: no-store`. Errors use the OAuth
form `{"error", "error_description"}`.
#### indieauth.com
- User authentication
- Authorization consent
- Authorization code generation
- Profile discovery
### Metadata
#### tokens.indieauth.com
- Token issuance
- Token verification
- Token revocation
- Scope management
## Configuration
### Required Settings
```ini
# Identity of the admin user
ADMIN_ME=https://your-domain.com
# External token endpoint for verification
TOKEN_ENDPOINT=https://tokens.indieauth.com/token
# Admin session secret (existing)
SECRET_KEY=your-secret-key
```
### HTML Discovery
```html
<!-- Added to all pages -->
<link rel="authorization_endpoint" href="https://indieauth.com/auth">
<link rel="token_endpoint" href="https://tokens.indieauth.com/token">
<link rel="micropub" href="https://starpunk.example.com/micropub">
```
## Security Model
### Trust Boundaries
```
┌─────────────────────────────────────────────────────────────┐
│ Trusted Zone │
│ ┌──────────────────────────────────────────────────────┐ │
│ │ StarPunk Application │ │
│ │ - Session management │ │
│ │ - Post creation/management │ │
│ │ - Admin interface │ │
│ └──────────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────┘
│
Token Verification API
│
┌─────────────────────────────────────────────────────────────┐
│ Semi-Trusted Zone │
│ ┌──────────────────────────────────────────────────────┐ │
│ │ External IndieAuth Providers │ │
│ │ - Token validation │ │
│ │ - Identity verification │ │
│ └──────────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────┘
│
User Authentication
│
┌─────────────────────────────────────────────────────────────┐
│ Untrusted Zone │
│ ┌──────────────────────────────────────────────────────┐ │
│ │ Micropub Clients │ │
│ │ - Must provide valid Bearer tokens │ │
│ │ - Tokens verified on every request │ │
│ └──────────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────┘
```
### Security Benefits of Simplified Architecture
1. **Reduced Attack Surface**
- No token generation = no cryptographic mistakes
- No token storage = no database leaks
- No PKCE = no implementation errors
2. **Specialized Security**
- Auth providers focus solely on security
- Regular updates from specialized teams
- Community-vetted implementations
3. **Clear Boundaries**
- StarPunk only verifies, never issues
- Single source of truth (external provider)
- No confused deputy problems
## Performance Characteristics
### Token Verification Performance
```
Without Cache:
┌──────────┐ 200-500ms ┌─────────────┐
│ Micropub ├───────────────────→│Token Endpoint│
└──────────┘ └─────────────┘
With Cache (95% hit rate):
┌──────────┐ <1ms ┌─────────────┐
│ Micropub ├───────────────────→│ Memory Cache │
└──────────┘ └─────────────┘
```
### Cache Strategy
```python
Cache Key: SHA256(token)
Cache Value: {
'me': 'https://user.com',
'client_id': 'https://client.com',
'scope': 'create update delete',
'expires_at': timestamp + 300 # 5 minutes
```json
{
"issuer": "<SITE_URL>/",
"authorization_endpoint": "<SITE_URL>/auth/authorize",
"token_endpoint": "<SITE_URL>/auth/token",
"revocation_endpoint": "<SITE_URL>/auth/revoke",
"revocation_endpoint_auth_methods_supported": ["none"],
"scopes_supported": ["create", "update", "delete", "media", "draft", "profile"],
"response_types_supported": ["code"],
"token_endpoint_auth_methods_supported": ["none"],
"client_id_metadata_document_supported": true,
"grant_types_supported": ["authorization_code"],
"code_challenge_methods_supported": ["S256"],
"authorization_response_iss_parameter_supported": true
}
```
### Expected Latencies
- First request: 200-500ms (external API)
- Cached request: <1ms
- Admin login: 1-2s (OAuth flow)
- Post creation: <50ms (after auth)
StarPunk has no introspection or userinfo endpoint. Only StarPunk itself
checks its tokens.
## Migration Impact
### Authorization request validation (`validate_authorization_request`)
### Breaking Changes
1. **All existing tokens invalid**
- Users must re-authenticate
- No migration path for tokens
- `client_id` must be an http(s) URL without a fragment.
- `redirect_uri` must be present and have no fragment. It must either share
the `client_id`'s scheme, host and port, or be published by the client. To
check the second case, StarPunk fetches the `client_id` URL (5 s timeout,
with an SSRF check on the URL unless `DEV_MODE` is on). It reads
`redirect_uris` from a JSON client metadata document, or
`rel="redirect_uri"` from the `Link` header and from `<link>`/`<a>`
elements.
- Problems with `client_id` or `redirect_uri` are shown to the user on an
error page (HTTP 400). They are never redirected, because the redirect
target can't be trusted.
- Other problems are redirected to the client with `error`,
`error_description`, `state` and `iss`. These are: a `response_type` other
than `code` (`unsupported_response_type`), a missing `state`, a missing
`code_challenge`, or a `code_challenge_method` other than `S256`.
- `scope` is split on spaces. Requested scopes are not filtered against
`scopes_supported`.
2. **Endpoint removal**
- `/auth/authorization` → 404
- `/auth/token` → 404
### Consent screen
3. **Configuration required**
- Must set `ADMIN_ME`
- Must configure domain with IndieAuth links
If `GET /auth/authorize` gets a valid request but there is no admin session,
StarPunk stores the request URL and redirects to the OIDC login, which
returns to that URL afterwards. With a session, it renders
`templates/admin/authorize.html`. The page shows the client id, the identity
the client will act as, the redirect URI, and one checkbox per requested
scope (all checked). A request without scopes is described as "only wants to
confirm your identity".
### Non-Breaking Preserved Functionality
1. **Admin login unchanged**
- Same URL (`/admin/login`)
- Same provider (IndieLogin.com)
- Sessions preserved
`POST /auth/authorize/approve` checks the session again and re-validates the
request from hidden form fields. "Deny" redirects with
`error=access_denied`. "Approve" grants only the scopes that were both
requested and left checked, issues a code, and redirects with `code`,
`state` and `iss`.
2. **Micropub API unchanged**
- Same endpoint (`/micropub`)
- Same request format
- Same response format
### Codes and tokens
## Comparison with Other Systems
| | Authorization code | Access token |
|---|---|---|
| Table | `indieauth_codes` | `indieauth_tokens` |
| Value | 32 random bytes, URL-safe | 32 random bytes, URL-safe |
| Stored as | SHA-256 hash | SHA-256 hash |
| Lifetime | 10 minutes | `INDIEAUTH_TOKEN_LIFETIME_DAYS` (default 90) |
| Bound to | `client_id`, `redirect_uri`, `scope`, `code_challenge`, `resource` | `client_id`, `scope`, `resource` |
### WordPress + IndieAuth Plugin
- **Similarity**: External provider for auth
- **Difference**: WP has user management, we don't
`redeem_authorization_code` handles redemption for both `POST /auth/token`
and `POST /auth/authorize`:
### Known IndieWeb Sites
- **micro.blog**: Custom auth server (complex)
- **Indigenous**: Client only, uses external auth
- **StarPunk**: Micropub server only (simple)
1. `code` and `code_verifier` are required.
2. An atomic `UPDATE ... SET used_at = now WHERE code_hash = ? AND used_at IS
NULL AND expires_at > now` makes the code single-use, even across
gunicorn workers.
3. `client_id` and `redirect_uri` must match the code exactly.
4. `BASE64URL(SHA256(code_verifier))` must equal the stored challenge
(constant-time comparison).
### Architecture Philosophy
```
"Do one thing well"
│
├── StarPunk: Publish notes
├── IndieAuth.com: Authenticate users
└── Tokens.indieauth.com: Manage tokens
The token endpoint accepts only `grant_type=authorization_code`. It refuses
a code issued with no scope (`invalid_grant`), because such codes are for
the profile-only flow at the authorization endpoint. The response is
`{access_token, token_type: "Bearer", scope, me, expires_in}`. There are no
refresh tokens: a client re-authorizes when its token expires.
Expired codes are deleted whenever a new code is issued.
**Resource indicators (v2.15.0, ADR-075).** An authorization request may
include `resource` (RFC 8707). The only accepted value is `SITE_URL/mcp`
(a trailing slash is ignored); anything else is `invalid_target`. With it,
a request without `scope` is offered `create update delete media draft`, and
the consent screen says the client is an AI assistant using MCP. The resource
is stored on the code and then on the token. A token request that names a
different resource gets `invalid_target`. Micropub clients send no resource
and their tokens have none. The MCP endpoint only accepts tokens for its
resource; see [mcp.md](mcp.md).
### Verification and revocation
- `verify_access_token(token)` hashes the token and accepts it if it is
neither revoked nor expired, then updates `last_used_at`. It returns
`{me: ADMIN_ME, client_id, scope, resource, expires_at}`. The legacy
`GET /auth/token` still answers with only `me`, `client_id` and `scope`.
- `check_scope(required, scope)` matches whole words in the space-separated
scope string.
- Revocation sets `revoked_at`. It can happen by token value, through
`POST /auth/revoke` or the legacy `POST /auth/token` with `action=revoke`
(unknown tokens are ignored, per RFC 7009). It can also happen by id, from
the admin page.
- `/admin/tokens` ("Connected Apps") lists active tokens with their client,
scope, dates and last use. Each has a Revoke button.
### Scopes
| Scope | Allows |
|---|---|
| `create` | Micropub create (published or draft), media upload, `q=source`, `q=category` |
| `draft` | Micropub create, always saved as a draft (v2.9.0); `q=source`, `q=category` |
| `update` | Micropub update; `q=source`, `q=category` |
| `delete` | Micropub delete and undelete; `q=source`, `q=category` |
| `media` | Media endpoint uploads |
| `profile` | Nothing beyond identity (there is no profile or userinfo endpoint) |
`q=config` and `q=syndicate-to` work with any valid token.
## 3. Token use at the Micropub and media endpoints
`routes/micropub.py` handles `GET/POST /micropub` and
`POST /micropub/media`. The old `POST /media` path is kept for existing
clients.
### Extracting the token (`micropub.extract_bearer_token`)
- The token comes from `Authorization: Bearer <token>`, with the scheme
matched case-insensitively (RFC 7235).
- Or it comes from `access_token` in the form body (POST) or the query
string (GET).
- A token in both places gets **400 `invalid_request`** (RFC 6750 §2,
v2.9.0).
- No token gets **401 `unauthorized`**, "No access token provided".
- An unknown, expired or revoked token gets **401 `unauthorized`**,
"Invalid or expired access token".
Every 401 carries `WWW-Authenticate: Bearer realm="micropub"`. When a token
was sent but isn't valid, `error="invalid_token"` is added (RFC 6750 §3,
v2.9.0).
### Scope rules
| Operation | Required scope | Notes |
|---|---|---|
| create | `create` or `draft` | With only `draft`, the post is saved unpublished whatever `post-status` says |
| update | `update` | |
| delete / undelete | `delete` | |
| `q=source`, `q=category` | any of `create`, `update`, `delete`, `draft` | These show drafts and tags (v2.9.0) |
| `q=config`, `q=syndicate-to` | any valid token | |
| media upload | `media` or `create` | `create` is accepted for older clients |
A missing scope gets **403 `insufficient_scope`** (RFC 6750). The Micropub
Recommendation's error table lists 401 for this case. StarPunk keeps 403 on
purpose (see the [v2.9.0 review](../design/v2.9.0/2026-11-03-micropub-review.md),
"Kept as is").
Micropub request handling itself is described in
[indieweb-protocols.md](indieweb-protocols.md#micropub).
## 4. Discovery
`templates/base.html` puts these links on every page, whether or not
`WEBMENTION_ENABLED` is set (v2.8.0):
```html
<link rel="micropub" href="<SITE_URL>/micropub">
<link rel="indieauth-metadata" href="<SITE_URL>/.well-known/oauth-authorization-server">
<link rel="authorization_endpoint" href="<SITE_URL>/auth/authorize">
<link rel="token_endpoint" href="<SITE_URL>/auth/token">
<link rel="me" href="<author home>">
<link rel="author" href="<author home>">
```
## Future Considerations
Micropub clients start from the URL the user types, which is normally
`ADMIN_ME`. If `ADMIN_ME` is the StarPunk site itself, the links above are
enough. If it is another page, that page must advertise StarPunk's
`indieauth-metadata` (or `authorization_endpoint` and `token_endpoint`) and
`micropub` links. Otherwise clients won't find StarPunk
([ADR-064](../decisions/ADR-064-oidc-admin-login-and-builtin-indieauth-server.md) §3).
### Potential V2 Enhancements (NOT for V1)
1. **Multi-user support**
- Would require user management
- Still use external auth
## 5. Data (migration 011)
2. **Multiple token endpoints**
- Support different providers per user
- Endpoint discovery from user domain
| Table | Purpose | Key columns |
|---|---|---|
| `sessions` | Admin browser sessions | `session_token_hash`, `me`, `expires_at`, `last_used_at`, `user_agent`, `ip_address` |
| `oidc_login_state` | One pending OIDC login | `state` (PK), `nonce`, `code_verifier`, `next_url`, `expires_at` |
| `indieauth_codes` | Authorization codes | `code_hash`, `client_id`, `redirect_uri`, `scope`, `code_challenge`, `resource` (migration 022), `expires_at`, `used_at` |
| `indieauth_tokens` | Access tokens | `token_hash`, `client_id`, `scope`, `resource` (migration 022), `expires_at`, `last_used_at`, `revoked_at` |
3. **Token caching layer**
- Redis for distributed caching
- Longer TTL with refresh
The legacy v1 tables `auth_state`, `tokens` and `authorization_codes` are
not used (see the migration's header comment).
### Explicitly NOT Implementing
1. **Custom authorization server**
- Violates simplicity principle
- Maintenance burden
## 6. Security properties and known gaps
2. **Password authentication**
- Not IndieWeb compliant
- Security burden
In place:
3. **JWT validation**
- Not part of IndieAuth spec
- Unnecessary complexity
- No secrets at rest: session tokens, codes and access tokens are stored only
as SHA-256 hashes.
- OIDC: PKCE S256, a nonce, single-use expiring state, issuer checks (in the
metadata and the RFC 9207 `iss`), full ID-token validation, and a single
allowed `sub`.
- IndieAuth: PKCE S256 required, single-use 10-minute codes bound to the
client and redirect URI, `iss` in every authorization response, and
untrusted redirect URIs never followed.
- Post-login redirects are limited to this site.
## Testing Strategy
CSRF (v2.11.0): admin form POSTs, `/auth/logout` and the consent form's
approve button need a `csrf_token` field (or `X-CSRF-Token` header) equal to
an HMAC-SHA256 of the `starpunk_session` cookie under `SECRET_KEY`; anything
else gets 400. The token changes with every login and needs no storage.
Endpoints apps call with bearer tokens are not affected. The client metadata
fetch uses `webmention_ssrf.safe_get` (every redirect hop checked, the
connection pinned, a 5-second limit).
### Unit Tests
```python
# Test external verification
@patch('httpx.get')
def test_token_verification(mock_get):
# Mock successful response
mock_get.return_value.status_code = 200
mock_get.return_value.json.return_value = {
'me': 'https://example.com',
'scope': 'create'
}
Known gaps:
result = verify_token('test-token')
assert result is not None
```
### Integration Tests
```python
# Test with real endpoint (in CI)
def test_real_token_verification():
# Use test token from tokens.indieauth.com
token = get_test_token()
result = verify_token(token)
assert result['me'] == TEST_USER
```
### Manual Testing
1. Configure domain with IndieAuth links
2. Use Quill or Indigenous
3. Create test post
4. Verify token caching
## Metrics for Success
### Quantitative Metrics
- **Code removed**: >500 lines
- **Database tables removed**: 2
- **Complexity reduction**: ~40%
- **Test coverage maintained**: >90%
- **Performance**: <500ms token verification
### Qualitative Metrics
- **Clarity**: Clear separation of concerns
- **Maintainability**: No auth code to maintain
- **Security**: Specialized providers
- **Flexibility**: User choice of providers
- **Simplicity**: Focus on core functionality
---
**Document Version**: 1.0
**Created**: 2025-11-24
**Author**: StarPunk Architecture Team
**Purpose**: Document simplified authentication architecture after IndieAuth server removal
- Unknown scopes are shown and granted as requested; they just have no
effect. (`draft` has been listed in `scopes_supported` since v2.15.0.)
- No token introspection endpoint and no refresh tokens.
+267 -190
View File
@@ -1,233 +1,310 @@
# Syndication Architecture
# Syndication (POSSE) Architecture
## Overview
StarPunk's syndication architecture provides multiple feed formats for content distribution, ensuring broad compatibility with feed readers and IndieWeb tools while maintaining simplicity.
Current as of v2.14.0. Code: `starpunk/syndication.py`, `starpunk/backfeed.py`.
## Current State (v1.1.0)
```
┌─────────────┐
│ Database │
│ (Notes) │
└──────┬──────┘
│
┌──────▼──────┐
│ feed.py │
│ (RSS 2.0) │
└──────┬──────┘
│
┌──────▼──────┐
│ /feed.xml │
│ endpoint │
└─────────────┘
```
StarPunk publishes on its own site first and then copies each note to other
networks (POSSE: Publish on your Own Site, Syndicate Elsewhere). Later
edits, unpublishing and deletes follow the copies.
## Target Architecture (v1.1.2+)
```
┌─────────────┐
│ Database │
│ (Notes) │
└──────┬──────┘
│
┌──────▼──────────────────┐
│ Feed Generation Layer │
├──────────┬───────────────┤
│ feed.py │ json_feed.py │
│ RSS/ATOM│ JSON │
└──────────┴───────────────┘
│
┌──────▼──────────────────┐
│ Feed Endpoints │
├─────────┬───────────────┤
│/feed.xml│ /feed.atom │
│ (RSS) │ (ATOM) │
├─────────┼───────────────┤
│ /feed.json │
│ (JSON Feed) │
└─────────────────────────┘
```
> Feeds (RSS, Atom, JSON Feed, OPML) and WebSub are pull and push
> distribution of the site itself, not crossposting. They are described in
> [indieweb-protocols.md](indieweb-protocols.md#feeds). This document
> covered feeds before v2.1.0.
## Design Principles
Decisions: [ADR-065](../decisions/ADR-065-direct-syndication-posse.md)
(GoToSocial, queue model), [ADR-066](../decisions/ADR-066-bluesky-syndication.md)
(Bluesky), [ADR-068](../decisions/ADR-068-likes-and-reposts.md) (likes and
reposts), [ADR-069](../decisions/ADR-069-threads-syndication.md) (Threads),
[ADR-070](../decisions/ADR-070-crosspost-links-and-cards.md) (links and link
cards), [ADR-073](../decisions/ADR-073-backfeed.md) (backfeed),
[ADR-074](../decisions/ADR-074-notifications.md) (notifications, including
crossposting problems), and the [v2.9.0 Micropub review](../design/v2.9.0/2026-11-03-micropub-review.md)
(bookmarks, `mp-syndicate-to` default).
### 1. Format Independence
Each syndication format operates independently:
- No shared state between formats
- Failures in one don't affect others
- Can be enabled/disabled individually
## Targets
### 2. Shared Data Access
All formats read from the same data source:
- Single query pattern for notes
- Consistent ordering (newest first)
- Same publication status filtering
A target exists when its credentials are configured. `get_targets()` builds
the list from config on each call and caches one client object per target
in the process.
### 3. Library Leverage
Maximize use of existing libraries:
- `feedgen` for RSS and ATOM
- Native Python `json` for JSON Feed
- No custom XML generation
| Target id | Client | Enabled by | Micropub `uid` | Text limit | Images | Edits follow? |
|---|---|---|---|---|---|---|
| `gotosocial` | `MastodonClient` (any Mastodon-API server) | `GOTOSOCIAL_URL` + `GOTOSOCIAL_ACCESS_TOKEN` | `GOTOSOCIAL_URL/` | `GOTOSOCIAL_MAX_CHARS` (5000) | `GOTOSOCIAL_MAX_MEDIA` (4) | Yes |
| `bluesky` | `BlueskyClient` (AT Protocol, app password) | `BLUESKY_HANDLE` + `BLUESKY_APP_PASSWORD` (`BLUESKY_SERVICE`, default `https://bsky.social`) | `https://bsky.app/profile/<handle>` | `BLUESKY_MAX_CHARS` (300) | 4 | No |
| `threads` | `ThreadsClient` (Meta Graph API) | `THREADS_ACCESS_TOKEN` (`THREADS_USERNAME` for links) | `https://www.threads.com/@<username>` | `THREADS_MAX_CHARS` (500) | `THREADS_MAX_MEDIA` (4) | No |
## Component Design
Display names come from `GOTOSOCIAL_NAME`, `BLUESKY_NAME` and
`THREADS_NAME`. GoToSocial statuses use `GOTOSOCIAL_VISIBILITY` (default
`public`). The GoToSocial token needs `write:statuses` and `write:media`,
plus `read:search` and `write:favourites` for likes and boosts, and
`read:statuses` and `read:accounts` (or `read`) for backfeed. Backfeed on
Threads needs `threads_read_replies` and `threads_manage_insights`. The
Threads client uses the API host `graph.threads.com` (v2.13.0).
### Feed Generation Module (`feed.py`)
**Current Responsibility**: RSS 2.0 generation
**Future Enhancement**: Add ATOM generation function
Bluesky and Threads can't edit posts through their APIs. Once a copy exists
there, edits to the note don't change it, but unpublish and delete still
remove it.
```python
# Pseudocode structure
def generate_rss_feed(notes, config) -> str
def generate_atom_feed(notes, config) -> str # New
```
## Choosing targets
### JSON Feed Module (`json_feed.py`)
**New Component**: Dedicated JSON Feed generation
`select_targets(note_id, target_ids)` records which targets a note goes to.
Every configured target not in the list is disabled; if it already has a
copy, that copy is deleted. Targets the note never went to get no row.
```python
# Pseudocode structure
def generate_json_feed(notes, config) -> str
def format_json_item(note) -> dict
```
| Origin | Targets |
|---|---|
| Admin editor, new note | One checkbox per target, all checked by default |
| Admin editor, edit | Checkboxes reflect the targets currently enabled; saving applies the selection |
| Micropub create | Only the targets in `mp-syndicate-to` (form field, `mp-syndicate-to[]`, or inside JSON `properties`). Values may be a target's `uid` (trailing slash ignored) or its id. **No `mp-syndicate-to` means no crossposting** (v2.9.0; before, all targets) |
| Micropub update / delete | Target selection is unchanged; the change follows existing copies |
### Route Handlers
Simple pass-through to generation functions:
```python
@app.route('/feed.xml') # Existing
@app.route('/feed.atom') # New
@app.route('/feed.json') # New
```
Micropub clients learn the targets from `q=syndicate-to` and `q=config`
(`[{"uid", "name"}]`). Selecting targets for a draft records them; nothing
is sent until the note is published.
## Data Flow
## Data: the `syndications` table (migration 012)
1. **Request**: Client requests feed at endpoint
2. **Query**: Fetch published notes from database
3. **Transform**: Convert notes to format-specific structure
4. **Serialize**: Generate final output (XML/JSON)
5. **Response**: Return with appropriate Content-Type
Migration 020 adds the backfeed columns; see [Backfeed](#backfeed-v2130-adr-073)
and [database.md](database.md#crossposting).
## Microformats2 Architecture
One row per (note, target), unique on `(note_id, target)`:
### Template Layer Enhancement
Microformats2 operates at the HTML template layer:
| Column | Meaning |
|---|---|
| `enabled` | Whether the note should have a copy on this target |
| `remote_id`, `remote_url` | The copy: status id, `at://` URI, Threads media id, or a `favourite:<id>` / `reblog:<id>` marker for native likes and boosts |
| `state` | `pending`, `synced` or `failed` |
| `version` | Incremented on every change that needs a reconcile |
| `attempts`, `last_error`, `next_attempt_at` | Retry bookkeeping |
| `claimed_until` | Lease held by the worker processing the row |
`note_id` deliberately has no foreign key, so the row outlives a hard
delete and the worker can still remove the copy.
## Change hooks
| Event | Call | Effect |
|---|---|---|
| Admin create/edit | `select_targets()` | Rows set to `pending`, version bumped |
| Micropub create | `select_targets(resolve_targets(mp-syndicate-to))` | After photos are attached |
| `update_note()` (admin or Micropub, including publish/unpublish) | `note_changed(id)` | Every row of the note set to `pending`, version bumped, attempts reset |
| Micropub photo changes | `note_changed(id)` | Same |
| `delete_note()` (soft or hard), `undelete_note()` | `note_changed(id)` | Same |
| Retry button on the edit page | `retry(id)` | That row set to `pending` now, attempts reset |
| Check now (`POST /admin/syndication/<id>/check`) | `backfeed.check_now(id)` | That copy's responses checked at the next worker pass |
| Backfill responses (`POST /admin/backfeed/backfill`) | `backfeed.start_backfill()` | Every eligible copy flagged for one check |
## Worker
`SyndicationWorker` is a daemon thread started in every gunicorn process
when at least one target is configured (not under `TESTING`). It runs
`process_due()` every `SYNDICATION_POLL_INTERVAL` seconds (default 10), and
loops again immediately while there is work.
`process_due()`:
1. Calls each client's `maintain()` (the Threads token refresh).
2. Selects up to 5 rows that are `pending`, due, and not claimed.
3. Claims each row with an atomic
`UPDATE ... SET claimed_until = now + 5 min WHERE state = 'pending' AND
(claimed_until IS NULL OR claimed_until < now)`. The other processes skip
it. Unlike the Webmention and WebSub workers, there is no file lock: the
row claim is the only coordination.
4. Reconciles the row (below), then records the result.
### Reconcile
```
┌──────────────┐
│ Data Model │
│ (Notes) │
└──────┬───────┘
│
┌──────▼───────┐
│ Templates │
│ + mf2 markup│
└──────┬───────┘
│
┌──────▼───────┐
│ HTML Output │
│ (Semantic) │
└──────────────┘
note published and row enabled?
no → remote copy exists? delete it → (None, None)
yes → copy exists and target can't edit? → keep it as is
copy exists → update (if it's gone remotely: create again)
no copy → create
```
### Markup Strategy
- **Progressive Enhancement**: Add classes without changing structure
- **CSS Independence**: Use mf2-specific classes, not styling classes
- **Validation First**: Test with parsers during development
Deleted notes are not found by `get_note()`, so they count as "not
published".
## Configuration Requirements
### Versions
### New Configuration Variables
```ini
# Author information for h-card
AUTHOR_NAME = "Site Author"
AUTHOR_URL = "https://example.com"
AUTHOR_PHOTO = "/static/avatar.jpg" # Optional
The row's `version` is read when the row is picked up. `_finish()` sets
`state = 'synced'` only if the version is unchanged. If the note was edited
while the send was in flight, the row stays `pending` and is reconciled
again with the new content. `_fail()` does the same.
# Feed settings
FEED_LIMIT = 50
FEED_FORMATS = "rss,atom,json" # Comma-separated
```
### Retries and idempotency
## Performance Considerations
- `SyndicationError(permanent=True)` fails the row immediately. This covers
HTTP 4xx other than 429, a missing target, images too large for Bluesky,
a Threads container in `ERROR`/`EXPIRED`, and Graph API auth/permission
errors (codes 10, 190, 200).
- Other errors retry after `min(30 s × 2^attempts, 6 h)`. After 8 attempts
the row becomes `failed`. The error is shown on the edit page and on the
admin Log page (v2.18.0), both with a Retry button. When a row finally fails, the owner also gets a `problems`
notification (ADR-074).
- `RemoteGone` (404, or Graph error 100/33) during an update means the copy
was deleted remotely, so a new one is created. During a delete it counts as
success.
- Duplicate posts after a lost response are prevented as follows:
- GoToSocial: an `Idempotency-Key: starpunk-<row id>-<version>` header.
- Bluesky and Threads: on a retry (`attempts > 0`), the last 25 posts are
searched for one containing the permalink (for Threads reposts, the
reposted URL). A match is adopted instead of posting again.
- Threads' publish step sometimes fails transiently ("Fatal"); it is
retried unless the error is an auth error.
### Caching Strategy
- Feed generation is read-heavy, write-light
- Consider caching generated feeds (5-minute TTL)
- Invalidate cache on note creation/update
## Composing posts
### Resource Usage
- RSS/ATOM: ~O(n) memory for n notes
- JSON Feed: Similar memory profile
- Microformats2: No additional server resources
`Post` carries the note, permalink, tags, the target's character limit,
media files, and for likes, reposts and bookmarks, the URL they point at and
its fetched title (from the reply-context cache).
## Security Considerations
### Text
### Content Sanitization
- HTML in feeds must be properly escaped
- CDATA wrapping for RSS/ATOM
- JSON string encoding for JSON Feed
- No script injection vectors
`compose_text()` builds: body, a blank line, the permalink, a blank line,
the hashtags. If that is too long, the body is cut at a word boundary with
"…"; the permalink always survives. Hashtags (`#` plus the tag's display
name without non-word characters) may take at most a third of the limit. A
bookmark's body starts with `🔖 <bookmarked URL>` (v2.9.0). A photo-only
note starts with the permalink.
### Rate Limiting
- Apply same limits as HTML endpoints
- Consider aggressive caching for feeds
- Monitor for feed polling abuse
| Target | Body format | Links | Hashtags |
|---|---|---|---|
| GoToSocial | The note's Markdown, sent with `content_type: text/markdown`; embedded images removed from the text (sent as attachments) | As written | In the text |
| Bluesky | Plain text from the rendered Markdown | Markdown link text becomes a clickable link facet; bare URLs and `#tags` get link/tag facets (UTF-8 byte offsets) | In the text, as tag facets |
| Threads | Plain text | Markdown links are written `text (url)`, up to 4 different URLs plus the permalink (Threads allows 5); the length counts non-ASCII characters as UTF-8 bytes | None in the text; the first usable tag becomes the post's `topic_tag` |
## Testing Architecture
### Link cards (ADR-070)
### Unit Tests
```
tests/
├── test_feed.py # Enhanced for ATOM
├── test_json_feed.py # New test module
└── test_microformats.py # Template parsing tests
```
- **Bluesky**: if the note has images, they are the embed. Otherwise, if
the note links somewhere, StarPunk builds an `app.bsky.embed.external`
card for the first link. `page_card()` fetches the page (at most 512 KB,
every redirect hop checked against private addresses, 10 s per request,
5 redirects) and reads the Open Graph, Twitter or HTML title, description
and image. The image (at most 5 MB) is shrunk under 1 MB and uploaded as
the thumbnail. If the page can't be fetched, the card has just the URL.
A note without links gets a card for its own permalink, built from the
note without fetching anything.
- **Threads**: no explicit `link_attachment` (Meta rejects one for a page
it hasn't crawled yet). Threads previews the first link in the text by
itself.
- **GoToSocial**: the server makes its own previews.
### Integration Tests
- Validate against external validators
- Test feed reader compatibility
- Verify IndieWeb tool parsing
### Media
## Backwards Compatibility
`_media_files()` collects the note's attached photos first, then images
embedded in the note text that are in this site's media library (some
Micropub clients put uploaded photos in the content). It stops at the
target's image limit. The `large` variant is used when one exists. Alt text
comes from the caption, or from the embedded image's alt.
### URL Structure
- `/feed.xml` remains RSS 2.0 (no breaking change)
- New endpoints are additive only
- Auto-discovery links updated in templates
| Target | Handling |
|---|---|
| GoToSocial | Uploaded via `POST /api/v2/media` with description; re-uploaded on each edit |
| Bluesky | `uploadBlob`; images over 1,000,000 bytes are re-encoded as JPEG at decreasing size and quality, and fail permanently if they still don't fit. Aspect ratio is sent |
| Threads | Meta fetches each image from its public `/media/` URL. Only JPEG and PNG up to 8 MB are used; others are left off. 1 image → `IMAGE` container, 2 or more → `CAROUSEL` of child containers. Alt text up to 1000 characters |
### Database
- No schema changes required
- All features use existing Note model
- No migration needed
Only notes of kind `note` carry media; likes, reposts and bookmarks don't.
## Future Extensibility
### Likes, reposts and bookmarks
### Potential Enhancements
1. Content negotiation on `/feed`
2. WebSub (PubSubHubbub) support
3. Custom feed filtering (by tag, date)
4. Feed pagination for large sites
| Kind | GoToSocial | Bluesky | Threads |
|---|---|---|---|
| Like of a post on that network | Native favourite (`favourite:<id>`) | Native `app.bsky.feed.like` | Not crossposted (the API can't like) |
| Like of anything else | Not crossposted | Not crossposted | Not crossposted |
| Repost without comment, of a post on that network | Native boost (`reblog:<id>`) | Native `app.bsky.feed.repost` | Not native: posts `🔁 title` + URL |
| Repost with comment, of a post on that network | Status: comment, `🔁 title`, URL | Quote post (`app.bsky.embed.record`) | Comment, `🔁 title`, URL |
| Repost of another page | Status: (comment,) `🔁 title`, URL | Post with an external card for the URL | (Comment,) `🔁 title`, URL |
| Bookmark | Ordinary status starting `🔖 <url>` | Ordinary post starting `🔖 <url>` | Ordinary post starting `🔖 <url>` |
### Format Support Matrix
| Format | v1.1.0 | v1.1.2 | v1.2.0 |
|--------|--------|--------|--------|
| RSS 2.0 | ✅ | ✅ | ✅ |
| ATOM | ❌ | ✅ | ✅ |
| JSON Feed | ❌ | ✅ | ✅ |
| Microformats2 | Partial | Partial | ✅ |
"On that network" is decided by the client. GoToSocial resolves the URL
through `/api/v2/search?resolve=true`. Bluesky matches
`https://bsky.app/profile/<actor>/post/<rkey>` and resolves it to a URI and
CID. Native likes and boosts have no URL of their own and nothing to edit.
Deleting the note undoes them (`unfavourite` / `unreblog`, or deleting the
like or repost record). A like that isn't crossposted is recorded as synced
with no remote copy.
## Decision Rationale
## u-syndication
### Why Multiple Formats?
1. **No Universal Standard**: Different ecosystems prefer different formats
2. **Low Maintenance**: Feed formats are stable, rarely change
3. **User Choice**: Let users pick their preferred format
4. **IndieWeb Philosophy**: Embrace plurality and interoperability
`templates/note.html` lists every row that has a `remote_url` as "Also on
<name>", using `<a class="u-syndication" rel="syndication">`. Micropub
`q=source` returns the same URLs as the `syndication` property. The
Webmention worker rejects a webmention whose source is one of the note's
own copies (e.g. backfed by Bridgy), so a copy is never shown as a response
to itself (v2.8.0).
### Why This Architecture?
1. **Simplicity**: Each component has single responsibility
2. **Testability**: Isolated components are easier to test
3. **Maintainability**: Changes to one format don't affect others
4. **Performance**: Can optimize each format independently
## Backfeed (v2.13.0, ADR-073)
## References
- [RSS 2.0 Specification](https://www.rssboard.org/rss-specification)
- [ATOM RFC 4287](https://tools.ietf.org/html/rfc4287)
- [JSON Feed Specification](https://www.jsonfeed.org/)
- [Microformats2](https://microformats.org/wiki/microformats2)
`starpunk/backfeed.py` brings responses to the copies back to the note.
After each crosspost pass, the worker loop calls `backfeed.process_due()`,
which does nothing unless `BACKFEED_ENABLED`; otherwise it checks up to three synced copies of published notes (not likes or
reposts) that are due, each under a lease (`backfeed_claimed_until`):
- **Reading.** Each client has `responses(remote_id, remote_url)`, which
returns a `Harvest`: GoToSocial `favourited_by`, `reblogged_by` and
public direct replies from `context`; Bluesky `getLikes`,
`getRepostedBy`, `getQuotes` and `getPostThread` (depth 1), plus
`getProfiles` for reply and quote authors' bios; Threads `replies` and
`insights` (like, repost and quote counts). Lists are paged up to 1,000
items; a list cut short makes the harvest incomplete.
- **Storing.** Responses are upserted into `webmentions` with `origin` (the
target id) and `remote_id`, `validation_status = 'valid'`, keyed by
`(source_url, target_url)`. Likes and reposts use the copy's URL plus
`#liked-by-<actor>` / `#reposted-by-<actor>`. A new row gets the
`WEBMENTION_AUTO_APPROVE` default; an existing row keeps its moderation.
A complete harvest marks that silo's rows for the note that weren't
found `invalid` ("No longer on <silo>"). Authors with `#nobridge` or
`#nobot` in their bio, or Bluesky's `!no-unauthenticated` label, are not
stored and their rows are deleted; since v2.13.3 their likes, reposts and
quotes are counted anonymously instead (`hidden_counts()`). If the note's
shown responses changed, its webmentions are re-sent (spec 3.1.4).
`syndications.likes_count`, `reposts_count` and `quotes_count` hold the
silo-only counts: all of Threads; elsewhere people who opted out. New
responses found by a backfill, or by a copy's first check on a note older
than a day, are stored as already notified (v2.14.0).
- **Schedule.** `backfeed_next_at` follows the note's age: 15 minutes (day
1), 1 hour (week 1), 1 day (to 30 days), 1 week (to
`BACKFEED_MAX_AGE_DAYS`), with up to 10% jitter; none after. Failures back
off like crossposts; after 8 they wait a month.
- **Backfill.** `start_backfill()` flags every eligible copy
(`backfeed_backfill = 1`). Flagged copies are checked only when no
routine check is due, at most one every 5 seconds across workers
(`site_state.backfeed_backfill_last`). Start and finish times are in
`site_state`.
- **Refused credentials.** A 401/403 (or Graph auth error) puts the target
in `backfeed.denied` until restart; the admin Log page says which
scopes are needed, and a `problems` notification is sent. Crossposting
is unaffected.
Avatars of all responses (received and backfed) are saved by
`starpunk/avatars.py` and served from `/avatars/<sha256>.webp`; the
webmention worker fills in missing ones every 10 minutes and removes unused
ones weekly.
## Client sessions and tokens
- **Bluesky**: `createSession` with the app password, then
`refreshSession` when the access JWT expires (a 401 or `ExpiredToken`
triggers one re-login and retry). The PDS endpoint is taken from the DID
document. Calls are serialized per client with a lock.
- **Threads**: long-lived tokens expire after 60 days. `maintain()` tries a
refresh (`refresh_access_token`, `th_refresh_token`) at most once a day.
The current token is kept in `DATA_PATH/threads_token.json` (mode 0600,
written atomically), together with a SHA-256 "seed" of the configured
token. A different `THREADS_ACCESS_TOKEN` in the config therefore
replaces the saved one. An exclusive `flock` on `threads_token.lock` makes
one gunicorn process refresh while the others pick up its result. A failed
refresh is retried an hour later (network error) or the next day (HTTP
error, e.g. a token less than a day old). The httpx logger is raised to
WARNING because Threads takes the token in the URL, and errors never echo
the URL.
- **GoToSocial**: a static bearer token.
## Known limits
- Bluesky and Threads copies don't follow edits (platform limitation).
Unticking and re-ticking a target makes a new copy.
- No per-target choice for Micropub updates: `mp-syndicate-to` is read only
on create.
- Likes of ordinary web pages, and all likes on Threads, stay on the site.
File diff suppressed because it is too large. Load diff
@@ -1,7 +1,9 @@
# ADR-005: IndieLogin Authentication Integration
## Status
Accepted
Superseded by ADR-064 (was: Accepted)
> Admin login stopped using IndieLogin.com during v1.x (it discovered the authorization endpoint from `ADMIN_ME`) and has used OIDC since v2.0.0 (ADR-064). (status updated 2026-11-03)
## Context
The user has explicitly required external IndieLogin authentication via indielogin.com for V1. This is different from implementing a full IndieAuth server (which CLAUDE.MD mentions). The distinction is important:
@@ -1,7 +1,9 @@
# ADR-007: Slug Generation Algorithm
## Status
Accepted
Superseded by ADR-062 (was: Accepted)
> New notes get timestamp slugs by default since v1.5.0 (ADR-062). (status updated 2026-11-03)
## Context
@@ -466,7 +468,7 @@ def update_note_slug(note_id, new_slug):
- [IndieWeb Permalink Design](https://indieweb.org/permalink)
- [URL Slug Best Practices](https://moz.com/learn/seo/url)
- [Python secrets Module](https://docs.python.org/3/library/secrets.html)
- [ADR-004: File-Based Note Storage](/home/phil/Projects/starpunk/docs/decisions/ADR-004-file-based-note-storage.md)
- [ADR-004: File-Based Note Storage](ADR-004-file-based-note-storage.md)
## Acceptance Criteria
@@ -444,9 +444,9 @@ When ready to release:
### Internal Documentation
- [Versioning Strategy](/home/phil/Projects/starpunk/docs/standards/versioning-strategy.md) - Complete versioning specification
- [Architecture Overview](/home/phil/Projects/starpunk/docs/architecture/overview.md) - System architecture
- [Development Setup](/home/phil/Projects/starpunk/docs/standards/development-setup.md) - Development workflow
- [Versioning Strategy](../standards/versioning-strategy.md) - Complete versioning specification
- [Architecture Overview](../architecture/overview.md) - System architecture
- [Development Setup](../standards/development-setup.md) - Development workflow
---
@@ -470,10 +470,10 @@ Created comprehensive documentation:
### Internal Documentation
- [ADR-008: Versioning Strategy](/home/phil/Projects/starpunk/docs/decisions/ADR-008-versioning-strategy.md) - Semantic versioning decision
- [Versioning Strategy](/home/phil/Projects/starpunk/docs/standards/versioning-strategy.md) - Complete versioning spec
- [Git Branching Strategy](/home/phil/Projects/starpunk/docs/standards/git-branching-strategy.md) - Complete branching spec
- [Development Setup](/home/phil/Projects/starpunk/docs/standards/development-setup.md) - Development workflow
- [ADR-008: Versioning Strategy](ADR-008-versioning-strategy.md) - Semantic versioning decision
- [Versioning Strategy](../standards/versioning-strategy.md) - Complete versioning spec
- [Git Branching Strategy](../standards/git-branching-strategy.md) - Complete branching spec
- [Development Setup](../standards/development-setup.md) - Development workflow
---
@@ -3,6 +3,8 @@
## Status
Accepted
> Partly superseded by ADR-064: the IndieLogin login flow was replaced by OIDC. The single `starpunk/auth.py` module and the session design are still in use. (status updated 2026-11-03)
## Context
With the core utilities and notes management complete, StarPunk needs an authentication system for the admin interface. ADR-005 already decided on using IndieLogin.com as the authentication provider. This ADR defines the specific module architecture and implementation approach for the authentication system.
@@ -204,7 +206,7 @@ Balance between security and usability:
## References
- [ADR-005: IndieLogin Authentication](/home/phil/Projects/starpunk/docs/decisions/ADR-005-indielogin-authentication.md)
- [ADR-005: IndieLogin Authentication](ADR-005-indielogin-authentication.md)
- [IndieAuth Specification](https://www.w3.org/TR/indieauth/)
- [OWASP Session Management](https://cheatsheetseries.owasp.org/cheatsheets/Session_Management_Cheat_Sheet.html)
- [Flask Security Best Practices](https://flask.palletsprojects.com/en/3.0.x/security/)
@@ -507,8 +507,8 @@ Dev auth implementation is successful if:
## References
- [ADR-005: IndieLogin Authentication](/home/phil/Projects/starpunk/docs/decisions/ADR-005-indielogin-authentication.md)
- [ADR-010: Authentication Module Design](/home/phil/Projects/starpunk/docs/decisions/ADR-010-authentication-module-design.md)
- [ADR-005: IndieLogin Authentication](ADR-005-indielogin-authentication.md)
- [ADR-010: Authentication Module Design](ADR-010-authentication-module-design.md)
- [OWASP Authentication Cheat Sheet](https://cheatsheetseries.owasp.org/cheatsheets/Authentication_Cheat_Sheet.html)
- [The Twelve-Factor App - Dev/Prod Parity](https://12factor.net/dev-prod-parity)
@@ -0,0 +1,181 @@
# ADR-012: Flaky Test Removal and Test Quality Standards
## Status
Accepted (implemented in v1.5.0; was: Proposed)
## Context
The test suite contains several categories of flaky tests that pass inconsistently. These tests consume developer time without providing proportional value. Per the project philosophy ("Every line of code must justify its existence"), we must evaluate whether these tests should be kept, fixed, or removed.
## Analysis by Test Category
### 1. Migration Race Condition Tests (`test_migration_race_condition.py`)
**Failing Tests:**
- `test_debug_level_for_early_retries` - Log message matching
- `test_new_connection_per_retry` - Connection count assertions
- `test_concurrent_workers_barrier_sync` - Multiprocessing pickle errors
- `test_sequential_worker_startup` - Missing table errors
- `test_worker_late_arrival` - Missing table errors
- `test_single_worker_performance` - Missing table errors
- `test_concurrent_workers_performance` - Pickle errors
**Value Analysis:**
- The migration retry logic with exponential backoff is *critical* for production deployments with multiple Gunicorn workers
- However, the flaky tests are testing implementation details (log levels, exact connection counts) rather than behavior
- The multiprocessing tests fundamentally cannot work reliably because:
1. `multiprocessing.Manager().Barrier()` objects cannot be pickled for `Pool.map()`
2. The worker functions require Flask app context that doesn't transfer across processes
3. SQLite database files in temp directories may not be accessible across process boundaries
**Root Cause:** Test design is flawed. These are attempting integration/stress tests using unit test infrastructure.
**Recommendation: REMOVE the multiprocessing tests entirely. KEEP and FIX the unit tests.**
Specifically:
- **REMOVE:** `TestConcurrentExecution` class (all 3 tests) - fundamentally broken by design
- **REMOVE:** `TestPerformance` class (both tests) - same multiprocessing issues
- **KEEP:** `TestRetryLogic` - valuable, just needs mock fixes
- **KEEP:** `TestGraduatedLogging` - valuable, needs logger configuration fixes
- **KEEP:** `TestConnectionManagement` - valuable, needs assertion fixes
- **KEEP:** `TestErrorHandling` - valuable, tests critical rollback behavior
- **KEEP:** `TestBeginImmediateTransaction` - valuable, tests locking mechanism
**Rationale for removal:** If we need to test concurrent migration behavior, that requires:
1. A proper integration test framework (not pytest unit tests)
2. External process spawning (not multiprocessing.Pool)
3. Real filesystem isolation
4. This is out of scope for V1 - the code works; the tests are the problem
---
### 2. Feed Route Tests (`test_routes_feeds.py`)
**Failing Assertions:**
- Tests checking for exact `<?xml version="1.0"` but code produces `<?xml version='1.0'` (single quotes)
- Tests checking for exact Content-Type with charset but response may vary
- Tests checking for exact `<rss version="2.0"` string
**Value Analysis:**
- These tests ARE valuable - they verify feed output format
- The tests are NOT flaky per se; they are *brittle* due to over-specific assertions
**Root Cause:** Tests are asserting implementation details (quote style) rather than semantics (valid XML).
**Recommendation: FIX by loosening assertions**
Current (brittle):
```python
assert b'<?xml version="1.0"' in response.data
```
Better (semantic):
```python
assert b'<?xml version=' in response.data
assert b"encoding=" in response.data
```
The test file already has SOME tests using the correct pattern (lines 72, 103, 265). The ATOM test on line 84 is the outlier - it should match the RSS tests.
---
### 3. Feed Streaming Test (`test_routes_feed.py`)
**Failing Test:** `test_feed_route_streaming`
**Current assertion (line 124):**
```python
assert "ETag" in response.headers
```
**But the test comment says:**
```python
# Cached responses include ETags for conditional requests
# (Phase 3 caching was added, replacing streaming for better performance)
```
**Value Analysis:**
- The test title says "streaming" but the implementation uses caching
- The test is actually correct (ETag SHOULD be present)
- If ETag is NOT present, that's a bug in the feed caching implementation
**Root Cause:** This is not a flaky test - if it fails, there's an actual bug. The test name is misleading but the assertion is correct.
**Recommendation: KEEP and RENAME to `test_feed_route_caching`**
---
### 4. Search Security Tests (`test_search_security.py`)
**Analysis of the file:** After reviewing, I see no obviously flaky tests. All tests are:
- Testing XSS prevention (correct)
- Testing SQL injection prevention (correct)
- Testing input validation (correct)
- Testing pagination limits (correct)
**Possible flakiness sources:**
- FTS5 special character handling varies by SQLite version
- Tests that accept multiple status codes (200, 400, 500) are defensive, not flaky
**Recommendation: KEEP all tests**
If specific flakiness is identified, it's likely due to SQLite FTS5 version differences, which should be documented rather than the tests removed.
---
## Decision
### Remove Entirely
1. `TestConcurrentExecution` class from `test_migration_race_condition.py`
2. `TestPerformance` class from `test_migration_race_condition.py`
### Fix Tests (Developer Action Items)
1. **`test_routes_feeds.py` line 84:** Change `assert b'<?xml version="1.0"'` to `assert b'<?xml version='`
2. **`test_routes_feed.py` line 117:** Rename test from `test_feed_route_streaming` to `test_feed_route_caching`
3. **`test_migration_race_condition.py`:** Fix logger configuration in `TestGraduatedLogging` tests to ensure DEBUG level is captured
4. **`test_migration_race_condition.py`:** Fix mock setup in `test_new_connection_per_retry` to accurately count connection attempts
### Keep As-Is
1. All tests in `test_search_security.py`
2. All non-multiprocessing tests in `test_migration_race_condition.py` (after fixes)
3. All other tests in `test_routes_feeds.py` and `test_routes_feed.py`
## Rationale
1. **Project philosophy alignment:** Tests that cannot reliably pass do not justify their existence. They waste developer time and erode confidence in the test suite.
2. **Pragmatic approach:** The migration concurrency code is tested in production by virtue of running with multiple Gunicorn workers. Manual testing during deployment is more reliable than broken multiprocessing tests.
3. **Test semantics over implementation:** Tests should verify behavior, not implementation details like quote styles in XML.
4. **Maintainability:** A smaller, reliable test suite is better than a larger, flaky one.
## Consequences
### Positive
- Faster, more reliable CI/CD pipeline
- Increased developer confidence in test results
- Reduced time spent debugging test infrastructure
- Tests that fail actually indicate bugs
### Negative
- Reduced test coverage for concurrent migration scenarios
- Manual testing required for multi-worker deployments
### Mitigations
- Document the multi-worker testing procedure in deployment docs
- Consider adding integration tests in a separate test category (not run in CI) for concurrent scenarios
## Alternatives Considered
### 1. Fix the multiprocessing tests
**Rejected:** Would require significant refactoring to use subprocess spawning instead of multiprocessing.Pool. The complexity is not justified for V1 given the code works correctly in production.
### 2. Mark tests as `@pytest.mark.skip`
**Rejected:** Skipped tests are just noise. They either work and should run, or they don't work and should be removed. "Skip" is procrastination.
### 3. Use pytest-xdist for parallel testing
**Rejected:** Does not solve the fundamental issue of needing to spawn external processes with proper app context.
### 4. Move to integration test framework (e.g., testcontainers)
**Considered for future:** This is the correct long-term solution but is out of scope for V1. Should be considered for V2 if concurrent migration testing is deemed critical.
@@ -4,6 +4,10 @@
Accepted
> Its deferral of WebSub is superseded by ADR-071 (v2.10.0), and "no feed caching" by the feed cache added in v1.1.2. RSS is now one of three feed formats (ADR-038) and is served at `/feed.rss` (`/feed.xml` still works). (status updated 2026-11-03)
> Item titles: since v2.16.0 only articles (notes with a title) have one; notes have none, and no first-line or date title is made up (ADR-076).
## Context
Phase 5 requires implementing RSS feed generation for syndicating published notes. We need to decide on the implementation approach, feed format, caching strategy, and technical details for generating a standards-compliant RSS feed.
@@ -365,8 +369,8 @@ The decision will be validated by:
- [Feed Discovery](https://indieweb.org/feed_discovery)
### Internal Documentation
- [Architecture Overview](/home/phil/Projects/starpunk/docs/architecture/overview.md)
- [Phase 5 Design](/home/phil/Projects/starpunk/docs/designs/phase-5-rss-and-container.md)
- [Architecture Overview](../architecture/overview.md)
- [Phase 5 Design](../design/v1.0.0/phase-5-rss-and-container.md)
---
@@ -87,10 +87,10 @@ The developer should:
6. Tag release after merge: `git tag -a v0.6.0 -m "Release 0.6.0: RSS feed and production container"`
## References
- [Versioning Strategy](/home/phil/Projects/starpunk/docs/standards/versioning-strategy.md)
- [Git Branching Strategy](/home/phil/Projects/starpunk/docs/standards/git-branching-strategy.md)
- [Phase 5 Design](/home/phil/Projects/starpunk/docs/designs/phase-5-rss-and-container.md)
- [Phase 5 Quick Reference](/home/phil/Projects/starpunk/docs/designs/phase-5-quick-reference.md)
- [Versioning Strategy](../standards/versioning-strategy.md)
- [Git Branching Strategy](../standards/git-branching-strategy.md)
- [Phase 5 Design](../design/v1.0.0/phase-5-rss-and-container.md)
- [Phase 5 Quick Reference](../design/v1.0.0/phase-5-quick-reference.md)
---
@@ -4,6 +4,8 @@
**Superseded by ADR-019** - IndieLogin.com does not use h-app microformats for client discovery. PKCE implementation is the correct solution.
> ADR-019 was itself replaced by ADR-025, and the whole IndieLogin login flow by OIDC in ADR-064. (status updated 2026-11-03)
## Context
StarPunk uses IndieLogin.com as a delegated IndieAuth provider for admin authentication. During the first production deployment to https://starpunk.thesatelliteoflove.com, authentication failed with the error:
@@ -4,6 +4,8 @@
**Superseded by ADR-019** - IndieLogin.com does not require OAuth metadata endpoint. PKCE implementation is the correct solution.
> ADR-019 was itself replaced by ADR-025, and the whole IndieLogin login flow by OIDC in ADR-064. (status updated 2026-11-03)
## Context
StarPunk continues to experience "client_id is not registered" errors from IndieLogin.com despite implementing h-app microformats in ADR-016 and making them visible in ADR-006.
@@ -2,7 +2,9 @@
## Status
Accepted
Superseded by ADR-064 (was: Accepted)
> The IndieAuth login flow whose HTTP traffic this logged was replaced by OIDC in v2.0.0. General logging is described in ADR-054. (status updated 2026-11-03)
## Context
@@ -2,7 +2,9 @@
## Status
Proposed
Superseded by ADR-025 (was: Proposed)
> ADR-025 is the accepted version of this decision; its login flow was in turn replaced by OIDC (ADR-064). (status updated 2026-11-03)
## Context
@@ -1,7 +1,7 @@
# ADR-021: IndieAuth Provider Strategy
## Status
Accepted
Superseded by ADR-064 (was: Accepted)
## Context
@@ -525,8 +525,8 @@ If there is user demand for a more integrated solution, V2 could add:
- [IndieWeb Wiki: authorization-endpoint](https://indieweb.org/authorization-endpoint) - Endpoint details
### Related ADRs
- [ADR-005: IndieLogin Authentication](/home/phil/Projects/starpunk/docs/decisions/ADR-005-indielogin-authentication.md) - Original auth decision
- [ADR-010: Authentication Module Design](/home/phil/Projects/starpunk/docs/decisions/ADR-010-authentication-module-design.md) - Auth module structure
- [ADR-005: IndieLogin Authentication](ADR-005-indielogin-authentication.md) - Original auth decision
- [ADR-010: Authentication Module Design](ADR-010-authentication-module-design.md) - Auth module structure
### Community Examples
- [Aaron Parecki's IndieAuth Notes](https://aaronparecki.com/2025/10/08/4/cimd) - Client ID metadata adoption
@@ -1,7 +1,9 @@
# ADR-022: Fix IndieAuth Callback Route Mismatch
## Status
Proposed
Accepted (implemented in v0.9.2; was: Proposed)
> The `/auth` prefix remains; since v2.0.0 `/auth/callback` is the OIDC callback (ADR-064). (status updated 2026-11-03)
## Context
We have discovered a critical routing mismatch in our IndieAuth implementation that causes a 404 error when IndieAuth providers redirect back to our application.
@@ -1,7 +1,9 @@
# ADR-023: IndieAuth Client Identification Strategy
## Status
Accepted
Superseded by ADR-064 (was: Accepted)
> StarPunk is now a registered OIDC client for login and publishes no h-app. (status updated 2026-11-03)
## Context
@@ -1,7 +1,9 @@
# ADR-024: Static HTML Identity Pages for IndieAuth
## Status
Accepted
Superseded by ADR-064 (was: Accepted)
> The `ADMIN_ME` page must now advertise StarPunk's own IndieAuth endpoints, not indieauth.com's (ADR-064, decision 3). (status updated 2026-11-03)
## Context
@@ -2,7 +2,7 @@
## Status
Accepted
Superseded by ADR-064 (was: Accepted)
## Context
@@ -1,7 +1,7 @@
# ADR-026: IndieAuth Token Exchange Compliance
## Status
Accepted
Superseded by ADR-064 (was: Accepted)
## Context
StarPunk's IndieAuth implementation is failing to authenticate with certain providers (specifically gondulf.thesatelliteoflove.com) during the token exchange phase. The provider is rejecting our token exchange requests with a "missing grant_type" error.
@@ -1,7 +1,7 @@
# ADR-027: IndieAuth Authentication Endpoint Correction
## Status
Accepted
Superseded by ADR-064 (was: Accepted)
## Context
@@ -178,7 +178,7 @@ Continue current implementation, document incompatibility.
- [IndieAuth Specification Section 5.4](https://www.w3.org/TR/indieauth/#authentication-response): Authorization Code Verification for authentication flows
- [IndieAuth Specification Section 6.3](https://www.w3.org/TR/indieauth/#token-response): Token Endpoint for authorization flows
- [IndieAuth Authentication vs Authorization](https://indieweb.org/IndieAuth#Authentication_vs_Authorization): Community documentation
- [ADR-021: IndieAuth Provider Strategy](/home/phil/Projects/starpunk/docs/decisions/ADR-021-indieauth-provider-strategy.md): Related architectural decision
- [ADR-021: IndieAuth Provider Strategy](ADR-021-indieauth-provider-strategy.md): Related architectural decision
---
@@ -2,7 +2,9 @@
## Status
Proposed
Accepted (implemented in v1.0.0-rc.1; was: Proposed)
> Its token endpoint was removed by ADR-050; tokens are now issued and verified by StarPunk's own IndieAuth server (ADR-064). Update and delete were added by ADR-067. (status updated 2026-11-03)
## Context
@@ -1,7 +1,9 @@
# ADR-029: Micropub IndieAuth Integration Strategy
## Status
Accepted
Superseded by ADR-050, then ADR-064 (was: Accepted)
> The token endpoint designed here was removed in v1.0.0-rc.4 (ADR-050). v2.0.0 added a new built-in IndieAuth server (ADR-064). (status updated 2026-11-03)
## Context
@@ -1,7 +1,7 @@
# ADR-030: External Token Verification Architecture
## Status
Accepted
Superseded by ADR-043, then ADR-064 (was: Accepted)
## Context
@@ -1,7 +1,7 @@
# ADR-031: Database Migration System Redesign
## Status
Proposed
Accepted (implemented through ADR-032 and ADR-033; was: Proposed)
## Context
@@ -1,7 +1,7 @@
# ADR-033: Database Migration System Redesign
## Status
Proposed
Accepted (implemented in v1.1.0; was: Proposed)
## Context
The current migration system has a critical flaw: duplicate schema definitions exist between SCHEMA_SQL (used for fresh installs) and individual migration files. This violates the DRY principle and creates maintenance burden. When schema changes are made, developers must remember to update both locations, leading to potential inconsistencies.
+1 -1
View File
@@ -1,7 +1,7 @@
# ADR-034: Full-Text Search with SQLite FTS5
## Status
Proposed
Accepted (implemented in v1.1.0; was: Proposed)
## Context
Users need the ability to search through their notes efficiently. Currently, finding specific content requires manually browsing through notes or using external tools. A built-in search capability is essential for any content management system, especially as the number of notes grows.
+3 -1
View File
@@ -1,7 +1,9 @@
# ADR-035: Custom Slugs in Micropub
## Status
Proposed
Accepted (implemented in v1.1.0; was: Proposed)
> Default slugs later became timestamps (ADR-062); `mp-slug` is still honoured. (status updated 2026-11-03)
## Context
Currently, StarPunk auto-generates slugs from note content (first 5 words). While this works well for most cases, users may want to specify custom slugs for:
@@ -1,7 +1,9 @@
# ADR-036: IndieAuth Token Verification Method Diagnosis
## Status
Accepted
Superseded by ADR-064 (was: Accepted)
> Micropub tokens are verified locally since v2.0.0. (status updated 2026-11-03)
## Context
StarPunk is experiencing HTTP 405 Method Not Allowed errors when verifying tokens with the external IndieAuth provider (gondulf.thesatelliteoflove.com). The user questioned "why are we making GET requests to these endpoints?"
@@ -1,7 +1,7 @@
# ADR-022: Multiple Syndication Format Support
## Status
Proposed
Accepted (implemented in v1.1.2; was: Proposed)
## Context
StarPunk currently provides RSS 2.0 feed generation using the feedgen library. The IndieWeb community and modern feed readers increasingly support additional syndication formats:
@@ -1,7 +1,9 @@
# ADR-023: Strict Microformats2 Compliance
## Status
Proposed
Accepted (implemented in v1.3.0, reviewed in v2.8.0; was: Proposed)
> Since v2.16.0 `p-name` is the stored title and `p-summary` the author's summary or, in previews, the start of the text (ADR-076).
## Context
StarPunk currently implements basic microformats2 markup:
@@ -158,7 +158,7 @@ The lesson learned: Release candidates are valuable for discovering not just bug
- [Semantic Versioning 2.0.0](https://semver.org/)
- [ADR-008: Versioning Strategy](../standards/versioning-strategy.md)
- [ADR-021: IndieAuth Provider Strategy](./ADR-021-indieauth-provider-strategy.md)
- [StarPunk Philosophy](../architecture/philosophy.md)
- StarPunk Philosophy (`philosophy.md`, not preserved)
---
@@ -1,7 +1,9 @@
# ADR-043-CORRECTED: IndieAuth Endpoint Discovery Architecture
## Status
Accepted (Replaces incorrect understanding in previous ADR-030)
Superseded by ADR-064 (was: Accepted; it replaced ADR-030)
> Admin login uses OIDC and Micropub tokens are verified locally since v2.0.0; no endpoints are discovered from `ADMIN_ME`. (status updated 2026-11-03)
## Context
@@ -1,7 +1,7 @@
# ADR-031: IndieAuth Endpoint Discovery Implementation Details
## Status
Accepted
Superseded by ADR-064 (was: Accepted)
## Context
@@ -1,7 +1,9 @@
# ADR-050: Remove Custom IndieAuth Server
## Status
Proposed
Superseded by ADR-064 (was: Proposed; implemented in v1.0.0-rc.4)
> v2.0.0 brought back a built-in IndieAuth server for Micropub clients (ADR-064). (status updated 2026-11-03)
## Context
@@ -363,7 +365,7 @@ TOKEN_ENDPOINT=https://tokens.indieauth.com/token
- [IndieAuth Spec](https://www.w3.org/TR/indieauth/)
- [tokens.indieauth.com](https://tokens.indieauth.com/)
- [ADR-021: IndieAuth Provider Strategy](/home/phil/Projects/starpunk/docs/decisions/ADR-021-indieauth-provider-strategy.md)
- [ADR-021: IndieAuth Provider Strategy](ADR-021-indieauth-provider-strategy.md)
- [Micropub Spec](https://www.w3.org/TR/micropub/)
---
@@ -1,7 +1,9 @@
# ADR-053: Performance Monitoring Strategy
## Status
Accepted
Superseded by ADR-072 (was: Accepted, v1.1.1)
> The metrics system was removed in v2.12.0. (status updated 2026-11-03)
## Context
StarPunk v1.1.1 introduces performance monitoring to help operators understand system behavior in production. Currently, we have no visibility into:
@@ -0,0 +1,112 @@
# ADR-056: Use External IndieAuth Provider (Never Self-Host)
## Status
Superseded by ADR-064 (was: Accepted, "permanent, non-negotiable")
> In v2.0.0 the operator moved login to OIDC and StarPunk became its own IndieAuth server for Micropub clients (ADR-064). (status updated 2026-11-03)
## Context
StarPunk is a minimal IndieWeb CMS focused on **content creation and syndication**, not identity infrastructure. The project philosophy demands that every line of code must justify its existence.
The question of whether to implement self-hosted IndieAuth has been raised multiple times. This ADR documents the final, permanent decision on this matter.
## Decision
**StarPunk will NEVER implement self-hosted IndieAuth.**
We will always rely on external IndieAuth providers such as:
- indielogin.com (primary recommendation)
- Other established IndieAuth providers
This decision is **permanent and non-negotiable**.
## Rationale
### 1. Project Focus
StarPunk's mission is to be a minimal CMS for publishing IndieWeb content. Our core competencies are:
- Publishing notes with proper microformats
- Generating RSS/Atom/JSON feeds
- Implementing Micropub for content creation
- Media management for content
Identity infrastructure is explicitly **NOT** our focus.
### 2. Complexity vs Value
Implementing IndieAuth would require:
- OAuth 2.0 implementation
- Token management
- Security considerations
- Key storage and rotation
- User profile management
- Authorization code flows
This represents hundreds or thousands of lines of code that don't serve our core mission of content publishing.
### 3. Existing Solutions Work
External IndieAuth providers like indielogin.com:
- Are battle-tested
- Handle security updates
- Support multiple authentication methods
- Are free to use
- Align with IndieWeb principles of building on existing infrastructure
### 4. Philosophy Alignment
Our core philosophy states: "Every line of code must justify its existence. When in doubt, leave it out."
Self-hosted IndieAuth cannot justify its existence in a minimal content-focused CMS.
## Consequences
### Positive
- Dramatically reduced codebase complexity
- No security burden for identity management
- Faster development of content features
- Clear project boundaries
- User authentication "just works" via proven providers
### Negative
- Dependency on external service (indielogin.com)
- Cannot function without internet connection to auth provider
- No control over authentication user experience
### Mitigations
- Document clear setup instructions for using indielogin.com
- Support multiple external providers for redundancy
- Cache authentication tokens appropriately
## Alternatives Considered
### 1. Self-Hosted IndieAuth (REJECTED)
**Why considered:** Full control over authentication
**Why rejected:** Massive scope creep, violates project philosophy
### 2. No Authentication (REJECTED)
**Why considered:** Ultimate simplicity
**Why rejected:** Single-user system still needs access control
### 3. Basic Auth or Simple Password (REJECTED)
**Why considered:** Very simple to implement
**Why rejected:** Not IndieWeb compliant, poor user experience
### 4. Hybrid Approach (REJECTED)
**Why considered:** Optional self-hosted with external fallback
**Why rejected:** Maintains complexity we're trying to avoid
## Implementation Notes
All authentication code should:
1. Assume an external IndieAuth provider
2. Never include hooks or abstractions for self-hosting
3. Document indielogin.com as the recommended provider
4. Include clear error messages when auth provider is unavailable
## References
- Project Philosophy: "Every line of code must justify its existence"
- IndieAuth Specification: https://indieauth.spec.indieweb.org/
- indielogin.com: https://indielogin.com/
## Final Note
This decision has been made after extensive consideration and multiple discussions. It is final.
**Do not propose self-hosted IndieAuth in future architectural discussions.**
The goal of StarPunk is **content**, not **identity**.
@@ -0,0 +1,110 @@
# ADR-057: Media Attachment Model
## Status
Accepted
## Context
The v1.2.0 media upload feature needed a clear model for how media relates to notes. Initial design assumed inline markdown image insertion (like a blog editor), but user feedback clarified that notes are more like social media posts (tweets, Mastodon toots) where media is attached rather than inline.
Key insights from user:
- "Notes are more like tweets, thread posts, mastodon posts etc. where the media is inserted is kind of irrelevant"
- Media should appear at the TOP of notes when displayed
- Text content should appear BELOW media
- Multiple images per note should be supported
## Decision
We will implement a social media-style attachment model for media:
1. **Database Design**: Use a junction table (`note_media`) to associate media files with notes, allowing:
- Multiple media per note (max 4)
- Explicit ordering via `display_order` column
- Per-attachment metadata (captions)
- Future reuse of media across notes
2. **Display Model**: Media attachments appear at the TOP of notes:
- 1 image: Full width display
- 2 images: Side-by-side layout
- 3-4 images: Grid layout
- Text content always appears below media
3. **Syndication Strategy**:
- RSS: Embed media as HTML in description (universal support)
- ATOM: Use both `<link rel="enclosure">` and HTML content
- JSON Feed: Use native `attachments` array (cleanest)
4. **Microformats2**: Multiple `u-photo` properties for multi-photo posts
## Rationale
**Why attachment model over inline markdown?**
- Matches user mental model (social media posts)
- Simplifies UI/UX (no cursor tracking needed)
- Better syndication support (especially JSON Feed)
- Cleaner Microformats2 markup
- Consistent display across all contexts
**Why junction table over array column?**
- Better query performance for feeds
- Supports future media reuse
- Per-attachment metadata
- Explicit ordering control
- Standard relational design
**Why limit to 4 images?**
- Twitter limit is 4 images
- Mastodon limit is 4 images
- Prevents performance issues
- Maintains clean grid layouts
- Sufficient for microblogging use case
## Consequences
### Positive
- Clean separation of media and text content
- Familiar social media UX pattern
- Excellent syndication feed support
- Future-proof for media galleries
- Supports accessibility via captions
- Efficient database queries
### Negative
- No inline images in markdown content
- All media must appear at top
- Cannot mix text and images
- More complex database schema
- Additional JOIN queries needed
### Neutral
- Different from traditional blog CMSs
- Requires grid layout CSS
- Media upload is separate from text editing
## Alternatives Considered
### Alternative 1: Inline Markdown Images
Store media URLs in markdown content as `![alt](url)`.
- **Pros**: Traditional blog approach, flexible positioning
- **Cons**: Poor syndication, complex editing UX, inconsistent display
### Alternative 2: JSON Array in Notes Table
Store media IDs as JSON array column in notes table.
- **Pros**: Simpler schema, fewer tables
- **Cons**: Poor query performance, no per-media metadata, violates 1NF
### Alternative 3: Single Media Per Note
Restrict to one image per note.
- **Pros**: Simplest implementation
- **Cons**: Too limiting, doesn't match social media patterns
## Implementation Notes
1. Migration will create both `media` and `note_media` tables
2. Feed generators must query media via JOIN
3. Template must render media before content
4. Upload UI shows thumbnails, not markdown insertion
5. Consider lazy loading for performance
## References
- [IndieWeb multi-photo posts](https://indieweb.org/multi-photo)
- [Microformats2 u-photo property](https://microformats.org/wiki/h-entry#u-photo)
- [JSON Feed attachments](https://jsonfeed.org/version/1.1#attachments)
- [Twitter photo upload limits](https://help.twitter.com/en/using-twitter/tweeting-gifs-and-pictures)
@@ -0,0 +1,183 @@
# ADR-058: Image Optimization Strategy
## Status
Accepted
## Context
The v1.2.0 media upload feature requires decisions about image size limits, optimization, and validation. Based on user requirements:
- 4 images maximum per note (confirmed)
- No drag-and-drop reordering needed (display order is upload order)
- Image optimization desired
- Optional caption field for each image (accessibility)
Research was conducted on:
- Web image best practices (2024)
- IndieWeb implementation patterns
- Python image processing libraries
- Storage implications for single-user CMS
## Decision
### Image Limits
We will enforce the following limits:
1. **Count**: Maximum 4 images per note
2. **File Size**: Maximum 10MB per image
3. **Dimensions**: Maximum 4096x4096 pixels
4. **Formats**: JPEG, PNG, GIF, WebP only
### Optimization Strategy
We will implement **automatic resizing on upload**:
1. **Resize Policy**:
- Images larger than 2048 pixels (longest edge) will be resized
- Aspect ratio will be preserved
- Original quality will be maintained (no aggressive compression)
- EXIF orientation will be corrected
2. **Rejection Policy**:
- Files over 10MB will be rejected (before optimization)
- Dimensions over 4096x4096 will be rejected
- Invalid formats will be rejected
- Corrupted files will be rejected
3. **Processing Library**: Use **Pillow** for image processing
### Database Schema Updates
Add caption field to `note_media` table:
```sql
CREATE TABLE note_media (
id INTEGER PRIMARY KEY,
note_id INTEGER NOT NULL,
media_id INTEGER NOT NULL,
display_order INTEGER NOT NULL DEFAULT 0,
caption TEXT, -- Optional caption for accessibility
created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
FOREIGN KEY (note_id) REFERENCES notes(id) ON DELETE CASCADE,
FOREIGN KEY (media_id) REFERENCES media(id) ON DELETE CASCADE,
UNIQUE(note_id, media_id)
);
```
## Rationale
### Why 10MB file size limit?
- Generous for high-quality photos from modern phones
- Prevents storage abuse on single-user instance
- Reasonable upload time even on slower connections
- Matches or exceeds most social platforms
### Why 4096x4096 max dimensions?
- Covers 16-megapixel images (4000x4000)
- Sufficient for 4K displays (3840x2160)
- Prevents memory issues during processing
- Larger than needed for web display
### Why resize to 2048px?
- Optimal balance between quality and performance
- Retina-ready (2x scaling on 1024px display)
- Significant file size reduction
- Matches common social media limits
- Preserves quality for most use cases
### Why Pillow over alternatives?
- De-facto standard for Python image processing
- Fastest for basic resize operations
- Minimal dependencies
- Well-documented and stable
- Sufficient for our needs (resize, format conversion, EXIF)
### Why automatic optimization?
- Better user experience (no manual intervention)
- Consistent output quality
- Storage efficiency
- Faster page loads
- Users still get good quality
### Why no thumbnail generation?
- Adds complexity for minimal benefit
- Modern browsers handle image scaling well
- Single-user CMS doesn't need CDN optimization
- Can be added later if needed
## Consequences
### Positive
- Automatic optimization improves performance
- Generous limits support high-quality photography
- Captions improve accessibility
- Storage usage remains reasonable
- Fast processing with Pillow
### Negative
- Users cannot upload raw/unprocessed images
- Some quality loss for images over 2048px
- No manual control over optimization
- Additional processing time on upload
### Neutral
- Requires Pillow dependency
- Images stored at single resolution
- No progressive enhancement (thumbnails)
## Alternatives Considered
### Alternative 1: No Optimization
Accept images as-is, no processing.
- **Pros**: Simpler, preserves originals
- **Cons**: Storage bloat, slow page loads, memory issues
### Alternative 2: Strict Limits (1MB, 1920x1080)
Match typical web recommendations.
- **Pros**: Optimal performance, minimal storage
- **Cons**: Too restrictive for photography, poor UX
### Alternative 3: Generate Multiple Sizes
Create thumbnail, medium, and full sizes.
- **Pros**: Optimal delivery, responsive images
- **Cons**: Complex implementation, 3x storage, overkill for single-user
### Alternative 4: Client-side Resizing
Resize in browser before upload.
- **Pros**: Reduces server load
- **Cons**: Inconsistent quality, browser limitations, poor UX
## Implementation Notes
1. **Validation Order**:
- Check file size (reject if >10MB)
- Check MIME type (accept only allowed formats)
- Load with Pillow (validates file integrity)
- Check dimensions (reject if >4096px)
- Resize if needed (>2048px)
- Save optimized version
2. **Error Messages**:
- "File too large. Maximum size is 10MB"
- "Invalid image format. Accepted: JPEG, PNG, GIF, WebP"
- "Image dimensions too large. Maximum is 4096x4096"
- "Image appears to be corrupted"
3. **Pillow Configuration**:
```python
# Preserve quality during resize
image.thumbnail((2048, 2048), Image.Resampling.LANCZOS)
# Correct EXIF orientation
ImageOps.exif_transpose(image)
# Save with original quality
image.save(output, quality=95, optimize=True)
```
4. **Caption Implementation**:
- Add caption field to upload form
- Store in `note_media.caption`
- Use as alt text in HTML
- Include in Microformats markup
## References
- [MDN Web Performance: Images](https://developer.mozilla.org/en-US/docs/Web/Performance/images)
- [Pillow Documentation](https://pillow.readthedocs.io/)
- [Web.dev Image Optimization](https://web.dev/fast/#optimize-your-images)
- [Twitter Image Specifications](https://developer.twitter.com/en/docs/twitter-api/v1/media/upload-media/uploading-media/media-best-practices)
@@ -0,0 +1,281 @@
# ADR-059: Full Feed Media Standardization (Option 3)
## Status
Proposed (For v1.3.0 Backlog)
## Context
StarPunk v1.2.0 introduced media attachments for notes (images). The initial implementation embeds media as HTML in feed description fields. Option 2 (implemented in v1.2.x) adds Media RSS extension elements and JSON Feed image fields for better feed reader compatibility.
This ADR documents Option 3: Full Standardization, which provides comprehensive media support across all syndication formats, including video, audio, and advanced features. This is planned for v1.3.0 or later.
## Decision
Document the scope of "Full Standardization" for feed media support to be implemented in a future release. This option goes beyond Option 2's basic Media RSS support to include:
1. **Complete Media RSS Specification Support**
2. **Podcast RSS Support (RSS 2.0 enclosures for audio)**
3. **Video Support**
4. **Multiple Image Sizes/Thumbnails**
5. **Full JSON Feed 1.1 Media Compliance**
## Scope of Full Standardization
### 1. Complete Media RSS Implementation
**Research Required**: Full Media RSS specification at https://www.rssboard.org/media-rss
**Elements to Implement**:
- `<media:content>` with full attribute support:
- `url` (required) - Direct URL to media file
- `fileSize` - Size in bytes
- `type` - MIME type
- `medium` - Type: "image", "audio", "video", "document", "executable"
- `isDefault` - Boolean for default rendition
- `expression` - "full", "sample", "nonstop"
- `bitrate` - Kilobits per second
- `framerate` - Frames per second (video)
- `samplingrate` - Samples per second (audio)
- `channels` - Audio channels
- `duration` - Seconds
- `height` / `width` - Dimensions in pixels
- `lang` - RFC-3066 language code
- `<media:group>` - Container for multiple renditions of same content
- `<media:thumbnail>` - Multiple sizes with url, width, height, time
- `<media:title>` - Media title (type="plain" or "html")
- `<media:description>` - Media description (type="plain" or "html")
- `<media:keywords>` - Comma-separated keywords
- `<media:category>` - Categorization with scheme attribute
- `<media:credit>` - Credit attribution with role and scheme
- `<media:copyright>` - Copyright information
- `<media:rating>` - Content rating (scheme-based)
- `<media:hash>` - MD5/SHA-1 hash for integrity
- `<media:player>` - Embeddable player URL
**Effort Estimate**: 8-12 hours
### 2. Podcast RSS Support
**Research Required**:
- Apple Podcast RSS specification
- Google Podcast RSS requirements
- Podcast Index namespace (podcast:)
**Elements to Implement**:
- iTunes namespace (`xmlns:itunes="http://www.itunes.com/dtds/podcast-1.0.dtd"`):
- `<itunes:summary>` - Episode summary
- `<itunes:duration>` - Audio duration (HH:MM:SS)
- `<itunes:image>` - Episode artwork
- `<itunes:explicit>` - Content rating
- `<itunes:episode>` - Episode number
- `<itunes:season>` - Season number
- `<itunes:episodeType>` - "full", "trailer", "bonus"
- `<itunes:author>` - Author name
- `<itunes:owner>` - Owner contact
- Standard RSS `<enclosure>` for audio:
- `url` - Direct audio file URL
- `length` - File size in bytes
- `type` - MIME type (audio/mpeg, audio/mp4, etc.)
**Database Changes**:
- Add `duration` column to `note_media` table
- Add `media_type` enum (image, audio, video)
- Consider `podcast_metadata` table for series-level data
**Effort Estimate**: 10-16 hours
### 3. Video Support
**Research Required**:
- Video hosting considerations (storage, bandwidth)
- Supported formats (mp4, webm, ogg)
- Transcoding requirements
- Poster image generation
**Implementation Scope**:
- Accept video uploads via Micropub media endpoint
- Generate poster thumbnails automatically
- Include in Media RSS with proper video attributes:
- `medium="video"`
- `framerate`, `duration`, `bitrate`
- Associated `<media:thumbnail>` for poster
- HTML5 `<video>` element in feed description
- Consider video hosting limits (file size, duration)
**Database Changes**:
- Video-specific metadata in `media` table
- Poster image path
- Transcoding status (if needed)
**Effort Estimate**: 16-24 hours (significant)
### 4. Multiple Image Sizes (Thumbnails)
**Research Required**:
- Responsive image best practices
- WebP generation
- srcset/sizes patterns
**Implementation Scope**:
- Generate multiple sizes on upload:
- Thumbnail: 150x150 (square crop)
- Small: 320px width
- Medium: 640px width
- Large: 1280px width
- Original: preserved
- Store all sizes in `media_variants` table
- Include in Media RSS:
```xml
<media:group>
<media:content url="large.jpg" isDefault="true" width="1280" />
<media:content url="medium.jpg" width="640" />
<media:content url="small.jpg" width="320" />
</media:group>
<media:thumbnail url="thumb.jpg" width="150" height="150" />
```
- JSON Feed: Use `image` for default, include variants in `_starpunk` extension
**Database Changes**:
- `media_variants` table: media_id, variant_type, path, width, height, size_bytes
- Add `has_variants` boolean to `media` table
**Effort Estimate**: 8-12 hours
### 5. Full JSON Feed 1.1 Media Compliance
**Research Required**: JSON Feed 1.1 specification for extensions
**Implementation Scope**:
- Top-level `image` field (URL of first image, per spec)
- Top-level `banner_image` if applicable
- Item-level `image` field (main/featured image)
- Item-level `banner_image` for posts with banners
- Complete `attachments` array:
```json
{
"url": "https://example.com/media/image.jpg",
"mime_type": "image/jpeg",
"title": "Image caption",
"size_in_bytes": 245760,
"duration_in_seconds": null
}
```
- Audio attachments with `duration_in_seconds`
- Video attachments (if supported)
**Effort Estimate**: 4-6 hours
### 6. ATOM Feed Media Extensions
**Research Required**:
- ATOM Media extension namespace
- `<link rel="enclosure">` best practices
**Implementation Scope**:
- `<link rel="enclosure">` for each media item
- `type` attribute with MIME type
- `length` attribute with file size
- `title` attribute with caption
- Consider `<link rel="related">` for thumbnails
**Effort Estimate**: 3-5 hours
## Total Effort Estimate
| Feature | Minimum | Maximum |
|---------|---------|---------|
| Complete Media RSS | 8 hours | 12 hours |
| Podcast RSS Support | 10 hours | 16 hours |
| Video Support | 16 hours | 24 hours |
| Multiple Image Sizes | 8 hours | 12 hours |
| JSON Feed Compliance | 4 hours | 6 hours |
| ATOM Extensions | 3 hours | 5 hours |
| **Total** | **49 hours** | **75 hours** |
**Note**: Video support is the most complex feature and could be deferred to v1.4.0 "Media" release.
## Prerequisites
Before implementing Full Standardization:
1. **Option 2 Complete**: Basic Media RSS and JSON Feed `image` field
2. **Image Optimization**: ADR-058 image optimization strategy implemented
3. **Media Storage Architecture**: Clear path for large file storage
4. **Test Infrastructure**: Feed validation tests in place
## Implementation Phases
### Phase A: Enhanced Image Support (v1.3.0)
- Multiple image sizes/thumbnails
- Full Media RSS for images
- Enhanced JSON Feed attachments
- **Effort**: 12-18 hours
### Phase B: Audio Support (v1.3.x or v1.4.0)
- Podcast RSS implementation
- Audio duration extraction
- iTunes namespace
- **Effort**: 10-16 hours
### Phase C: Video Support (v1.4.0 "Media")
- Video upload handling
- Poster generation
- Video in feeds
- **Effort**: 16-24 hours
## Consequences
### Positive
- Best-in-class feed reader compatibility
- Podcast distribution capability
- Video content support
- Professional media syndication
- Future-proof architecture
### Negative
- Significant implementation effort (50-75 hours total)
- Increased storage requirements
- More complex feed generation
- Processing overhead for image variants
- Larger codebase to maintain
### Neutral
- Aligns with media-focused v1.4.0 roadmap
- Phased implementation possible
- Optional features can be configuration-gated
## Alternatives Considered
### Alternative 1: Minimal Enhancement (Option 2 Only)
Just implement basic Media RSS and JSON Feed image field.
- **Pros**: Low effort, immediate benefit
- **Cons**: Misses podcast/video opportunity
### Alternative 2: Third-Party Media Service
Use external service (Cloudinary, etc.) for media processing.
- **Pros**: Offloads complexity
- **Cons**: External dependency, cost, data ownership concerns
### Alternative 3: Plugin Architecture
Make media support pluggable for advanced features.
- **Pros**: Keeps core simple
- **Cons**: Added architectural complexity
## References
- [Media RSS Specification](https://www.rssboard.org/media-rss)
- [JSON Feed 1.1 Specification](https://jsonfeed.org/version/1.1)
- [Apple Podcast RSS Requirements](https://podcasters.apple.com/support/823-podcast-requirements)
- [Podcast Index Namespace](https://github.com/Podcastindex-org/podcast-namespace)
- [RSS 2.0 Enclosure Specification](https://www.rssboard.org/rss-specification#ltenclosuregtSubelementOfLtitemgt)
- [ADR-057: Media Attachment Model](ADR-057-media-attachment-model.md)
- [ADR-058: Image Optimization Strategy](ADR-058-image-optimization-strategy.md)
## Decision
This ADR documents the scope of Full Standardization (Option 3) for the project backlog. Implementation should be scheduled for v1.3.0 and v1.4.0 releases according to the phased approach outlined above.
**Immediate Action**: Implement Option 2 (ADR-060) for v1.2.x release.
**Future Action**: Review and refine this scope when scheduling v1.3.0 work.
@@ -0,0 +1,105 @@
# ADR-060: Route Adapter for the Metrics Dashboard
## Status
Superseded by ADR-072 (was: Accepted, v1.1.1-rc.2, 2025-11-25)
> The metrics dashboard was removed in v2.12.0. (status updated 2026-11-03)
> Reconstructed on 2026-11-03 from code, changelog and related documents;
> the original was lost. The original
> (`ADR-060-production-hotfix-metrics-dashboard.md`) was deleted in the
> v1.1.1-rc.2 architectural review as "a bug fix, not an architectural
> decision", and its content was merged into
> `docs/design/v1.1.1/hotfix-v1.1.1-rc2-consolidated.md`. Code
> (`starpunk/routes/admin.py`) and CHANGELOG 1.1.1-rc.2 still cite it, so
> this record restores the number.
>
> Not to be confused with the "Option 2 (ADR-060)" for Media RSS mentioned
> in ADR-059: that ADR was planned but never written (see
> `docs/design/v1.2.0/2025-12-10-documentation-audit-v1.2.0.md`).
## Context
v1.1.1-rc.1 added a metrics dashboard. In production it returned 500:
```
jinja2.exceptions.UndefinedError: 'dict object' has no attribute 'database'
```
Two problems were found:
1. **Route conflict.** The notes dashboard (`/admin/`, endpoint
`admin.dashboard`) and the new metrics dashboard (`/admin/dashboard`)
were easily confused. This was fixed first by moving the metrics
dashboard to `/admin/metrics-dashboard` (the "ADR-022" route-naming fix
cited in the CHANGELOG; see Numbering below).
2. **Data contract mismatch** (the real cause). The template reads
`metrics.database.count` and `.avg/.min/.max`, but
`starpunk.monitoring.get_metrics_stats()` returns
`by_type.database.count` and `avg_duration_ms`, `min_duration_ms`,
`max_duration_ms`.
Either the monitoring module, the template (and its JavaScript), or the
route could change.
## Decision
Adapt the data **in the route handler** (presentation layer) and leave
both the monitoring module's API and the template unchanged.
- `transform_metrics_for_template()` in `starpunk/routes/admin.py`
flattens `by_type.<type>` to `<type>` for `database`, `http` and
`render`, renames `avg_duration_ms`/`min_duration_ms`/`max_duration_ms`
to `avg`/`min`/`max`, and keeps `total_count`, `max_size` and
`process_id`. Missing types get zero values.
- If collecting metrics fails, the route flashes a warning and renders
the same structure with zeros.
- The monitoring imports are defensive: if `starpunk.monitoring` or the
pool statistics are unavailable, stub functions return an error value
and the dashboard still renders.
- The metrics dashboard lives at `/admin/metrics-dashboard`
(`admin.metrics_dashboard`); `/admin/` (`admin.dashboard`) stays the
notes list, so existing `url_for("admin.dashboard")` redirects are
unaffected.
## Consequences
### Positive
- Minimal, low-risk hotfix: one route changed, no template or monitoring
changes.
- The monitoring module's output stays the same for the JSON endpoint
(`/admin/metrics`) and other consumers.
- The dashboard degrades instead of failing when metrics are missing.
### Negative
- Treats the symptom: the template and the monitoring module still
disagree about the data shape, and the adapter must be kept in step
with both (noted in `docs/design/v1.1.1/hotfix-v1.1.1-rc2-review.md`).
- Errors loading metrics are shown as zeros, which can hide a problem.
## Alternatives Considered
- **Change the monitoring module to return the template's shape**: breaks
its other consumers.
- **Change the template and its JavaScript**: larger change in a hotfix,
with more to test.
## Numbering
The CHANGELOG for 1.1.1-rc.2 says "See ADR-022 and ADR-060". The ADR-022
it meant was `ADR-022-admin-dashboard-route-conflict-hotfix.md`, deleted in
the same review; its number collided with the existing
[ADR-022](ADR-022-auth-route-prefix-fix.md) (IndieAuth callback route) and
it was not restored. Its decision (the route rename) is recorded above.
## References
- `starpunk/routes/admin.py`: `transform_metrics_for_template()`,
`metrics_dashboard()`
- `docs/design/v1.1.1/hotfix-v1.1.1-rc2-consolidated.md`
- `docs/design/v1.1.1/2025-11-25-hotfix-v1.1.1-rc.2-implementation.md`
- `docs/design/v1.1.1/hotfix-v1.1.1-rc2-review.md`
- ADR-053 (performance monitoring)
+111
View File
@@ -0,0 +1,111 @@
# ADR-061: Author Profile Discovery from IndieAuth
## Status
Accepted
## Context
StarPunk v1.2.0 requires Microformats2 compliance, including proper h-card author information in h-entries. The original design assumed author information would be configured via environment variables (AUTHOR_NAME, AUTHOR_PHOTO, etc.).
However, since StarPunk uses IndieAuth for authentication, and users authenticate with their domain/profile URL, we have an opportunity to discover author information directly from their IndieWeb profile rather than requiring manual configuration.
The user explicitly stated: "These should be retrieved from the logged in profile domain (rel me etc.)" when asked about author configuration.
## Decision
Implement automatic author profile discovery from the IndieAuth 'me' URL:
1. When a user logs in via IndieAuth, fetch their profile page
2. Parse h-card microformats and rel-me links from the profile
3. Cache this information in a new `author_profile` database table
4. Use discovered information in templates for Microformats2 markup
5. Provide fallback behavior when discovery fails
## Rationale
1. **IndieWeb Native**: Discovery from profile URLs is a core IndieWeb pattern
2. **DRY Principle**: Author already maintains their profile; no need to duplicate
3. **Dynamic Updates**: Profile changes are reflected on next login
4. **Standards-Based**: Uses existing h-card and rel-me specifications
5. **User Experience**: Zero configuration for author information
6. **Consistency**: Author info always matches their IndieWeb identity
## Consequences
### Positive
- No manual configuration of author information required
- Automatically stays in sync with user's profile
- Supports full IndieWeb identity model
- Works with any IndieAuth provider
- Discoverable rel-me links for identity verification
### Negative
- Requires network request during login (mitigated by caching)
- Depends on proper markup on user's profile page
- Additional database table required
- More complex than static configuration
- Parsing complexity for microformats
### Implementation Details
#### Database Schema
```sql
CREATE TABLE author_profile (
id INTEGER PRIMARY KEY,
me_url TEXT NOT NULL UNIQUE,
name TEXT,
photo TEXT,
bio TEXT,
rel_me_links TEXT, -- JSON array
discovered_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
updated_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP
);
```
#### Discovery Flow
1. User authenticates with IndieAuth
2. On successful login, trigger discovery
3. Fetch user's profile page (with timeout)
4. Parse h-card for: name, photo, bio
5. Parse rel-me links
6. Store in database with timestamp
7. Use cache for 7 days, refresh on login
#### Fallback Strategy
- If discovery fails during login, use cached data if available
- If no cache exists, use minimal defaults (domain as name)
- Never block login due to discovery failure
- Log failures for monitoring
## Alternatives Considered
### 1. Environment Variables (Original Design)
Static configuration via .env file
- ✅ Simple, no network requests
- ❌ Requires manual configuration
- ❌ Duplicates information already on profile
- ❌ Can become out of sync
### 2. Hybrid Approach
Environment variables with optional discovery
- ✅ Flexibility for both approaches
- ❌ More complex configuration
- ❌ Unclear which takes precedence
### 3. Discovery Only, No Cache
Fetch profile on every request
- ✅ Always up to date
- ❌ Performance impact
- ❌ Reliability issues
### 4. Static Import Tool
CLI command to import profile once
- ✅ No runtime discovery needed
- ❌ Manual process
- ❌ Can become stale
## Implementation Priority
High - Required for v1.2.0 Microformats2 compliance
## References
- https://microformats.org/wiki/h-card
- https://indieweb.org/rel-me
- https://indieweb.org/discovery
- W3C IndieAuth specification
@@ -0,0 +1,197 @@
# ADR-062: Timestamp-Based Slug Format
## Status
Accepted (Supersedes ADR-007). Articles take their slug from the title since v2.17.0 ([ADR-077](ADR-077-article-urls-and-title-slugs.md)).
## Context
ADR-007 established a content-based slug generation algorithm that extracts the first 5 words from note content to create URL slugs. While this approach provides readable, SEO-friendly URLs, it has drawbacks for a personal note-taking system:
1. **Privacy Concerns**: The slug reveals note content in the URL. Private thoughts or draft content become visible in URLs that may be shared or logged.
2. **SEO Irrelevance**: StarPunk is designed for personal IndieWeb notes, not public blogs. Notes are typically short-form content (similar to tweets or status updates) where SEO optimization provides no meaningful benefit.
3. **Unpredictable Slugs**: Users cannot predict what slug will be generated without examining their content carefully.
4. **Edge Case Handling**: Content-based slugs require complex fallback logic for short content, unicode-only content, or special characters.
5. **Collision Complexity**: Similar notes require random suffix generation (e.g., `hello-world-a7c9`), which undermines the readability goal.
The user has explicitly stated: "Notes don't need SEO."
## Decision
Change the default slug format from content-based to timestamp-based:
**New Default Format**: `YYYYMMDDHHMMSS`
- Example: `20251216143052`
- Compact, sortable, predictable
- 14 characters total
**Collision Handling**: Sequential numeric suffix
- First collision: `20251216143052-1`
- Second collision: `20251216143052-2`
- Simple, predictable, no randomness
**Custom Slugs**: Continue to support user-specified slugs via `mp-slug` Micropub property and the web UI custom slug field. When provided, custom slugs take precedence.
### Algorithm Specification
```python
def generate_slug(custom_slug: str = None, created_at: datetime = None) -> str:
"""Generate a URL-safe slug for a note.
Args:
custom_slug: User-provided custom slug (takes precedence if provided)
created_at: Note creation timestamp (defaults to now)
Returns:
URL-safe slug string
"""
if custom_slug:
return sanitize_slug(custom_slug)
# Default: timestamp-based
timestamp = (created_at or datetime.now()).strftime("%Y%m%d%H%M%S")
return ensure_unique_slug(timestamp)
def ensure_unique_slug(base_slug: str) -> str:
"""Ensure slug is unique, adding numeric suffix if needed."""
if not slug_exists(base_slug):
return base_slug
suffix = 1
while slug_exists(f"{base_slug}-{suffix}"):
suffix += 1
return f"{base_slug}-{suffix}"
```
### Examples
| Scenario | Generated Slug |
|----------|----------------|
| Normal note at 2:30:52 PM on Dec 16, 2025 | `20251216143052` |
| Second note in same second | `20251216143052-1` |
| Third note in same second | `20251216143052-2` |
| Note with custom slug "my-custom-slug" | `my-custom-slug` |
## Rationale
### Timestamp Format (Score: 9/10)
**Pros**:
- **Privacy**: URL reveals nothing about content
- **Predictability**: User knows exactly what slug format to expect
- **Sortability**: Chronological sorting by URL is possible
- **Simplicity**: No complex word extraction or normalization
- **Collision Rarity**: Same-second creation is rare; handled cleanly when it occurs
- **Compact**: 14 characters vs potentially 50+ for content-based
**Cons**:
- **Not Memorable**: `20251216143052` is harder to remember than `hello-world`
- **No SEO Value**: Search engines prefer descriptive URLs
**Why Cons Don't Matter**:
- StarPunk is for personal notes, not public SEO-optimized content
- Notes are accessed via UI, feeds, or bookmarks, not memorized URLs
- Users wanting memorable URLs can use custom slugs
### Sequential Suffix (Score: 10/10)
**Pros**:
- **Deterministic**: No randomness; same collision always gets same suffix
- **Simple**: No cryptographic random generation needed
- **Readable**: `-1`, `-2` are clear and obvious
- **Debuggable**: Easy to understand collision resolution
**Cons**:
- **Enumerable**: Sequential numbers could be probed
- Not a real security concern for note slugs
### Comparison with ADR-007 Approach
| Aspect | ADR-007 (Content-Based) | ADR-062 (Timestamp) |
|--------|------------------------|---------------------|
| Privacy | Reveals content | Reveals only time |
| Complexity | High (word extraction, normalization, unicode handling) | Low (strftime) |
| SEO | Good | None |
| Predictability | Low | High |
| Collision handling | Random suffix | Sequential suffix |
| Fallback cases | Many (short content, unicode, etc.) | None |
| Code lines | ~50 | ~15 |
## Consequences
### Positive
1. **Simplified Code**: Remove complex word extraction, unicode normalization, and multiple fallback paths
2. **Better Privacy**: Note content never appears in URLs
3. **Predictable Output**: Users always know what slug format to expect
4. **Fewer Edge Cases**: No special handling for short content, unicode, or special characters
5. **Cleaner Collisions**: Sequential suffixes are more intuitive than random strings
### Negative
1. **Migration**: Existing notes keep their content-based slugs (no migration needed)
2. **Not Human-Readable**: URLs don't describe content
3. **No SEO**: Search engines won't benefit from descriptive URLs
### Mitigations
**Human-Readable URLs**:
- Users wanting descriptive URLs can use custom slugs via `mp-slug`
- The web UI custom slug field remains available
- This is opt-in rather than default
**SEO**:
- IndieWeb notes are typically not SEO targets
- Content is in the page body where search engines can index it
- Microformats2 markup provides semantic meaning
## Backward Compatibility
- Existing notes retain their slugs (no data migration)
- New notes use timestamp format by default
- Custom slug functionality unchanged
- All existing URLs remain valid
## Testing Requirements
- Test default slug generation produces timestamp format
- Test collision handling with sequential suffixes
- Test custom slugs still take precedence
- Test edge case: multiple notes in same second
- Test reserved slug rejection still works
- Verify existing tests for custom slugs pass
## Implementation Notes
Changes required in:
- `starpunk/slug_utils.py`: Update `generate_slug()` function
- `starpunk/notes.py`: Remove content parameter from slug generation call
- Tests: Update expected slug formats
Estimated effort: Small (1-2 hours implementation, 1 hour testing)
## References
- ADR-007: Slug Generation Algorithm (Superseded by this ADR)
- ADR-035: Custom Slugs (Unchanged; complements this decision)
- IndieWeb Permalink Best Practices: https://indieweb.org/permalink
## Acceptance Criteria
- [ ] Default slugs use `YYYYMMDDHHMMSS` format
- [ ] Collision handling uses sequential suffix (`-1`, `-2`, etc.)
- [ ] Custom slugs via `mp-slug` continue to work
- [ ] Custom slugs via web UI continue to work
- [ ] Reserved slug validation unchanged
- [ ] Existing notes unaffected
- [ ] All tests pass
- [ ] Code complexity reduced
---
**Approved**: 2025-12-16
**Architect**: StarPunk Architect Agent
**Supersedes**: ADR-007 (Slug Generation Algorithm)
@@ -0,0 +1,293 @@
# ADR-063: Webmention Implementation
## Status
Accepted (v1.6.0; revised v2.6.8, v2.6.9, v2.7.0, v2.7.2, v2.8.0, v2.10.1)
> Reconstructed on 2026-11-03 from code, changelog and related documents;
> the original was lost. The v1.6.0 release was never committed: its code
> was recovered from the production image (`v1.6.0-rc.2`), but its tests,
> this ADR and the `docs/design/v1.6.0/` folder (`background-worker.md`,
> `api-contract.md`, `display-moderation.md`, `rate-limiting.md`) were not
> in the image. This document describes the design as it exists in
> v2.10.1. The numbering of the decisions below is new, except that
> Decision 9 (SSRF protection) keeps the number the code cites.
## Context
Webmention (W3C Recommendation, https://www.w3.org/TR/webmention/) is how
IndieWeb sites tell each other that one page links to another. It is what
makes replies, likes and reposts from other sites (and from Mastodon or
Bluesky via Bridgy) show up on a note, and what tells the sites a note
replies to or links to that it exists.
Both directions involve fetching URLs chosen by someone else, which is
slow, can fail, and can be abused:
- A receiver must fetch the source page to check that it really links to
the target. Doing that inside the request lets anyone make StarPunk
fetch arbitrary URLs synchronously, and ties up a server process.
- A sender must discover each target's endpoint and POST to it, with
retries when the other site is down.
- Any fetch can be pointed (directly, by redirect or by DNS) at private
addresses on the server's own network (SSRF).
- Content from other sites is shown on our pages (XSS).
StarPunk is a single-user site on SQLite, run under gunicorn with several
processes, with no job queue or message broker.
## Decision
### 1. Endpoint and discovery
- `POST /webmention` receives webmentions (form-encoded as the spec
requires; JSON is accepted as an extension);
`GET /webmention` returns a short description of the endpoint.
- HTML responses advertise it with a `Link: </webmention>; rel="webmention"`
header (added alongside other Link headers, not replacing them) and a
`<link rel="webmention">` in `base.html`.
- `WEBMENTION_ENABLED` (default true) turns the feature off.
### 2. Receiving is asynchronous
The endpoint does only cheap, local checks and then answers
**202 Accepted**; verification happens later in the background worker.
Synchronous checks, each answered with 400 on failure: `source` and
`target` are well-formed http(s) URLs of at most 2048 characters, they
differ, and the target is a **published note** on this site (host compared
case-insensitively and without default ports; `#fragment` dropped). Since
v2.7.0 mentions of the home page or other pages are refused instead of
being stored and never shown.
The mention is upserted into `webmentions` keyed by `(source_url,
target_url)` with `validation_status = 'pending'`, so a re-sent mention is
re-verified rather than duplicated.
### 3. Background worker
One daemon thread (`WebmentionWorker`) handles both incoming verification
and outgoing sending, using two SQLite tables as queues (`webmentions`,
`webmention_queue`). No external queue is introduced.
- Started by `create_app()` when `WEBMENTION_WORKER_ENABLED` is true (not
in tests). It polls every `WEBMENTION_POLL_INTERVAL` seconds (10),
handles up to `WEBMENTION_BATCH_SIZE` items (5) per pass, and backs off
`WEBMENTION_ERROR_BACKOFF` seconds (60) after an unexpected error.
- Each pass runs inside `app.app_context()` (v2.6.8: without it every
item failed with "Working outside of application context"). The thread
lives for the life of the process; it is not stopped on app-context
teardown (v2.6.8: that stopped it after the first request).
- **Only one process works the queues** (v2.6.9). Every gunicorn process
starts a worker, but a worker only processes after taking a
non-blocking `flock` on `<data>/webmention-worker.lock`. Others retry
each poll, so one takes over if the holder dies.
- On acquiring the lock it resets outgoing items stuck in `sending` for
more than 5 minutes (orphans from a crash).
- `/health?detailed=true` (authenticated) reports the worker's
statistics.
### 4. Verifying a received webmention
Per Recommendation 3.2.2 and 3.2.4:
- The source is fetched with an `Accept` header, following redirects
(Decision 9), and decoded in its declared character set.
- The source must link to the **exact** target URL, by media type: for
HTML any link attribute (resolved against the final URL after
redirects); for JSON any string value; for `text/*` a substring; other
types never match. Fragments are ignored (v2.7.0).
- The source is parsed with mf2py. The responding entry and its author are
found with the authorship algorithm (v2.8.0); the mention type is one of
`reply`, `like`, `repost`, `bookmark` or `mention`. The source's own
`u-url` and `dt-published` are stored (`source_post_url`,
`source_published`, migration 014) so the response links to and is
dated by the original post.
- If the source is gone (410, other 4xx) or no longer links, the mention
is removed from display. If it can't be reached right now (network
error, 5xx, temporary DNS failure), a previously valid mention stays up
(v2.7.0, v2.10.1).
- A webmention from one of the note's own crossposts (e.g. its GoToSocial
copy returned through Bridgy) is ignored (v2.8.0).
### 5. Moderation and display
- Valid mentions are approved automatically when `WEBMENTION_AUTO_APPROVE`
is true (the default), otherwise they wait in `pending`. The admin can
approve, hide or delete them at `/admin/webmentions` (single and bulk
actions).
- A re-sent mention keeps its moderation: hidden stays hidden, approved
stays approved (v2.7.0).
- Notes show only mentions that are both `valid` and `approved`, grouped
by type. Replies are marked up as `u-comment h-cite`; likes, reposts,
bookmarks and mentions as `u-like`, `u-repost`, `u-bookmark`,
`u-mention` h-cites, each with the author's h-card (v2.8.0).
### 6. Sending
- After a note is created, edited, deleted or restored, links are
extracted from its **rendered HTML**: `<a href>` only, excluding links
inside `<pre>` and `<code>`, the `in_reply_to` target first, then
document order, at most 10 current links per operation. Each target is
queued in `webmention_queue` (unique per source and target).
- On edit, every target sent to before is notified again, including links
the edit removed and an old reply target; on delete and undelete too
(Recommendation 3.1.4, 3.1.5; v2.7.0). Re-sends are not capped.
- When the responses a note displays change (a mention arrives, changes
or is removed, or the admin approves, hides or deletes one that was
shown), the note's previous webmentions are re-queued, as 3.1.4
recommends. An unchanged re-verification sends nothing, so two sites
that both do this don't ping-pong (v2.7.2).
- Endpoint discovery follows 3.1.2: the first `Link` header with a
`webmention` rel token (RFC 8288 parsing, several headers, any case),
otherwise the first `<link>` or `<a>` in document order; relative URLs
resolve against the final URL. The endpoint is discovered afresh every
time; there is no cache (v2.7.0).
- The webmention is a form-encoded POST. Any 2xx is success; redirects
are not followed (a redirected POST would become a GET). A target
without an endpoint is marked `no_endpoint`.
- Failures (network errors, 5xx, 429) are retried after 1 h, 6 h and
24 h; after the fourth failed attempt the item is `failed`.
- The User-Agent is `StarPunk/<version> Webmention (+site)`.
### 7. Deleted notes
A deleted note answers **410 Gone** with a "This post has been deleted"
page (soft-deleted notes, and hard-deleted notes that sent webmentions),
so receivers can remove the mention (3.1.5; v2.7.0).
### 8. Content safety (XSS)
- Snippets and reply content are sanitized with nh3, cut at 500
characters and sanitized again so no tag is left open (v2.7.0); the
admin "Mentions" list shows them with tags stripped.
- Author URLs and photos are kept only if they are absolute http(s) URLs;
otherwise the source URL is used as the author link and no photo is
shown (v2.6.9). mf2 values with `alt` are reduced to their URL.
### 9. SSRF protection
All outbound requests for webmentions (source fetches, endpoint discovery
and the POST) go through `starpunk/webmention_ssrf.py` (Recommendation
4.2, 4.3, 4.5):
- Only http(s). **Every** address a hostname resolves to (IPv4 and IPv6)
must be public: private, loopback, link-local, multicast, reserved,
unspecified, carrier-grade NAT (`100.64.0.0/10`), `0.0.0.0/8` and
IPv4-mapped forms are blocked.
- Redirects (at most 10) are followed by hand and **each hop** is checked
(v2.7.0).
- **DNS pinning** (v2.7.2): each request connects to the exact address
that passed the check, with the original host in the `Host` header and
as TLS SNI, so certificates are still verified against the hostname. A
rebinding DNS server can't answer differently the second time. IPv4 is
preferred when a host has both.
- Responses are limited to 1 MB, and each fetch has a total time limit of
10 s that a slow, byte-by-byte response can't extend (v2.10.1).
- `DEV_MODE` allows localhost for local testing.
The same guard is reused for IndieAuth client metadata fetches (ADR-064),
Micropub remote photos (v2.9.0) and WebSub callbacks and deliveries
(ADR-071).
### 10. Rate limiting
`POST /webmention` is rate-limited before any other work: by client IP
(`WEBMENTION_RATE_LIMIT_IP`, 60) and by source domain
(`WEBMENTION_RATE_LIMIT_DOMAIN`, 10) per `WEBMENTION_RATE_LIMIT_WINDOW`
(60 s), answering 429 with `Retry-After`. Counters are an in-memory
sliding-window log (`starpunk/rate_limiter.py`), so they are per server
process. Behind a reverse proxy, `TRUSTED_PROXY=true` makes the first
`X-Forwarded-For` address the client IP. `WEBMENTION_RATE_LIMIT_ENABLED`
turns it off.
`starpunk/rate_limiter.py` and `routes/webmention.py` cite this as
"ADR-064: Webmention Rate Limiting", a v1.6.0 ADR that was lost with this
one. The number ADR-064 was later reused for OIDC login
(ADR-064-oidc-admin-login-and-builtin-indieauth-server.md); the rate-limit
decision is recorded here instead.
### 11. Data (migrations 010 and 014)
- `notes.in_reply_to`
- `webmentions`: source, target, target note, type, author name/URL/photo,
snippet, `validation_status` (`pending`, `valid`, `invalid`, `error`),
`moderation_status` (`pending`, `approved`, `hidden`, `deleted`),
timestamps, and (migration 014) `source_post_url`, `source_published`
- `webmention_queue`: outgoing items with `status` (`pending`, `sending`,
`sent`, `failed`, `no_endpoint`), retry count and next retry time
- `reply_contexts`: cached title, snippet and author of pages notes reply
to
## Consequences
### Positive
- Receiving is cheap and can't be used to make StarPunk fetch URLs while
the sender waits.
- No new infrastructure: the queues are SQLite tables and the worker is a
thread, as for crossposting and WebSub.
- Outbound fetching is hardened once and reused by every other feature
that fetches user-supplied URLs.
- Follows every MUST and SHOULD of the Recommendation after v2.7.2.
### Negative
- Mentions appear only after the next worker pass (seconds), not
immediately.
- One worker per site: throughput is limited, which is fine for a
personal site.
- Rate limits are per process and reset on restart.
- `error` rows (unexpected failures) are not retried automatically.
- Endpoint discovery ignores HTTP cache headers (spec 5.3); targets are
only fetched when a note changes.
- The 410 page is a generic tombstone without microformats.
## Revisions
- **v2.6.8** — Worker runs inside an app context and is no longer stopped
on teardown; before this, no webmention was ever processed
(`docs/design/v2.6.8/2026-11-03-webmention-worker-app-context.md`).
- **v2.6.9** — Author photos with alt text; endpoint discovery crash
fixed and discovery errors retried; `flock` so only one process works
the queues; http(s)-only author links
(`docs/design/v2.6.9/2026-11-03-webmention-processing-fixes.md`).
- **v2.7.0** — Full review against the Recommendation: published notes
only, exact target match, RFC 8288 endpoint discovery, no discovery
cache, re-sends on edit/delete/undelete, 410 for deleted notes,
per-hop SSRF checks, size and time limits, moderation kept on re-send
(`docs/design/v2.7.0/2026-11-03-webmention-spec-compliance.md`).
- **v2.7.2** — DNS pinning; re-send when displayed responses change
(`docs/design/v2.7.2/2026-11-03-webmention-shoulds.md`).
- **v2.8.0** — Authorship algorithm, source URL and published time
(migration 014), response microformats, own crossposts ignored
(`docs/design/v2.8.0/2026-11-03-microformats2-review.md`).
- **v2.10.1** — Hard total time limit on fetches; temporary DNS failures
treated as transient.
## Alternatives Considered
- **Synchronous verification in the request**: simpler, but slow, ties
up a gunicorn process and turns the endpoint into a fetch-anything
proxy. The Recommendation allows 202 for exactly this.
- **External service (webmention.io)**: no code to run, but mentions
would live outside the site's database and moderation, against the
self-hosted approach.
- **A job queue (Celery/RQ with Redis)**: new infrastructure for a few
jobs a day.
- **Resolving and validating the host, then letting the HTTP client
connect normally**: open to DNS rebinding; replaced by pinning in
v2.7.2.
## References
- W3C Webmention: https://www.w3.org/TR/webmention/
- Code: `starpunk/routes/webmention.py`, `starpunk/webmention_worker.py`,
`starpunk/webmention_processor.py`, `starpunk/webmention_sender.py`,
`starpunk/webmention_ssrf.py`, `starpunk/webmention_queries.py`,
`starpunk/rate_limiter.py`, `migrations/010_add_webmentions.sql`,
`migrations/014_webmention_source_details.sql`
- Tests: `tests/test_webmention_spec.py`, `tests/test_webmention_worker.py`
- [IndieWeb protocols](../architecture/indieweb-protocols.md)
- ADR-064 (OIDC login; reuses the SSRF guard), ADR-071 (WebSub; same
guard and worker pattern)
@@ -0,0 +1,70 @@
# ADR-064: OIDC Admin Login and Built-in IndieAuth Server
## Status
Accepted (v2.0.0, 2026-10-31)
Supersedes the login parts of ADR-021, ADR-025, ADR-027 and ADR-043, and the
external token verification of ADR-030 and ADR-036.
## Context
Up to v1.6.0 StarPunk was an IndieAuth *client*. Admin login discovered an
authorization endpoint from `ADMIN_ME` and used the authentication-only
flow; Micropub verified bearer tokens by calling the token endpoint
discovered from `ADMIN_ME`. In practice both pointed at Gondulf, a separate
self-hosted IndieAuth server.
The operator now runs Pocket ID (OIDC, passkeys) for every other
self-hosted service and wants StarPunk to use it too. Pocket ID cannot act
as an IndieAuth server: IndieAuth clients identify themselves by URL
(`client_id` = their homepage) and are not pre-registered, while OIDC
clients must be registered with the provider.
Micropub apps such as Quill and Indigenous only speak IndieAuth, so moving
login to OIDC alone would end Micropub posting.
## Decision
1. **Admin login uses OIDC** (authorization code flow, PKCE S256, nonce).
StarPunk is a confidential client. The ID token's signature (provider
JWKS), `iss`, `aud`, `exp` and `nonce` are verified, and only the
configured `OIDC_ADMIN_SUB` may log in. A successful login creates the
same local session as before, for `ADMIN_ME`.
2. **StarPunk becomes its own IndieAuth server** for Micropub clients:
`/.well-known/oauth-authorization-server`, `/auth/authorize` (consent
screen, admin session required, so approval is protected by OIDC),
`/auth/token` and `/auth/revoke`. Only `response_type=code` with PKCE
S256 is supported. Tokens are stored as SHA-256 hashes and verified
locally by `/micropub` and `/media`.
3. **`ADMIN_ME` stays** as the site owner's identity URL: it is the author
h-card source and the `me` in issued tokens. That page must advertise
StarPunk's IndieAuth metadata (or `authorization_endpoint` /
`token_endpoint` links) so Micropub apps discover StarPunk.
4. **New tables** (`oidc_login_state`, `indieauth_codes`,
`indieauth_tokens`, migration 011). The legacy `auth_state`, `tokens`
and `authorization_codes` tables are not reused because migration 004
drops the last two after `INITIAL_SCHEMA_SQL` creates them.
5. Client `redirect_uri` must share the `client_id` origin, or be published
by the client (JSON `redirect_uris` or `rel="redirect_uri"`). The fetch
reuses the webmention SSRF guard.
## Consequences
- Gondulf is no longer needed and can be retired.
- Existing admin sessions stay valid (sessions are keyed by `ADMIN_ME`, which
does not change). Notes, media, tags and webmentions are untouched.
- Micropub apps must be re-authorized once, since their old tokens were
issued by Gondulf. Tokens are visible and revocable at `/admin/tokens`.
- Production requires `OIDC_ISSUER`, `OIDC_CLIENT_ID`,
`OIDC_CLIENT_SECRET`, `OIDC_ADMIN_SUB`. `INDIELOGIN_URL` and
`TOKEN_ENDPOINT` are ignored.
- New dependency: `PyJWT[crypto]`.
- Major version bump (2.0.0): configuration and endpoint discovery change
incompatibly.
## Alternatives Considered
- **Keep Gondulf for Micropub only**: two auth systems to run and patch.
- **Drop Micropub**: simplest, but loses posting from apps.
- **Manually created API tokens**: Quill and most apps require IndieAuth.
@@ -0,0 +1,45 @@
# ADR-065: Direct Syndication (POSSE) to GoToSocial
## Status
Accepted (v2.1.0, 2026-10-31)
## Context
StarPunk posts were copied to other networks by Echofeed reading the feed.
That gives no control per post, no record of where a post went, and edits
and deletes never follow. The operator runs GoToSocial locally and wants
StarPunk to post there directly, with Bluesky likely next.
## Decision
1. **Publish here, copy out (POSSE).** Each (note, target) has a row in
`syndications` (migration 012). Creating, editing, unpublishing or
deleting a note marks its rows pending; a background worker reconciles
the remote post: create, update (`PUT /api/v1/statuses/:id`) or delete.
2. **Defaults.** New notes go to every configured target unless unticked in
the editor or a Micropub client sends `mp-syndicate-to` (an explicit
list, possibly empty). Existing notes are never backfilled.
3. **Content.** The full note as Markdown (`content_type: text/markdown`),
then the permalink and the note's tags as hashtags. Longer than the
target's limit: shortened at a word boundary with the link kept.
Images are uploaded first (the `large` variant) with their captions as
alt text.
4. **Delivery.** Rows are claimed atomically so the four gunicorn workers'
threads never send twice; an `Idempotency-Key` guards retries; a row
`version` keeps edits made during a send. Network errors and 5xx/429
retry with exponential backoff (8 attempts); other 4xx fail at once.
A remote post deleted by hand is posted again on the next edit.
5. **Backlinks.** Synced URLs are shown on the note page as `u-syndication`
links, and in the editor with status, last error and a retry button.
6. **Targets** are config-driven. One kind exists, a Mastodon-API client
(`GOTOSOCIAL_URL`, `GOTOSOCIAL_ACCESS_TOKEN`); it also works for
Mastodon. A Bluesky target would add a second client behind the same
reconcile loop.
## Consequences
- Echofeed must stop posting to GoToSocial or posts appear twice.
- The token needs only `write:statuses write:media`.
- `syndications.note_id` has no foreign key so a hard-deleted note's
remote post can still be deleted.
@@ -0,0 +1,37 @@
# ADR-066: Crossposting to Bluesky
## Status
Accepted (v2.2.0, 2026-10-31)
## Context
ADR-065 added direct crossposting to GoToSocial. Bluesky was still fed by
Echofeed, with the same problems: no per-post control, no record of the
copy, deletes never follow.
## Decision
1. **Second client, same loop.** `BlueskyClient` sits behind the reconcile
loop and `syndications` table from ADR-065, as target id `bluesky`.
Clients share a small interface: `create`, `update`, `delete`, `can_edit`.
2. **Auth.** An app password (`BLUESKY_HANDLE`, `BLUESKY_APP_PASSWORD`).
`createSession` against `BLUESKY_SERVICE` (default `https://bsky.social`);
writes go to the PDS named in the returned DID document. Expired access
tokens are refreshed, falling back to a new session.
3. **Content.** Plain text (Markdown rendered and stripped), then the
permalink and hashtags (at most a third of the space), shortened to 300
characters at a word boundary. Link and hashtag facets use UTF-8 byte
offsets. Images (up to 4) are re-encoded as JPEG if over 1 MB, with alt
text and aspect ratio; a note without images gets a link card.
4. **Edits don't follow.** Bluesky has no edit; rewriting the record would
break likes and replies, and delete-and-repost would lose them. The post
stays as first sent. Unpublish, untick and delete remove it.
5. **No double posts.** When a create is retried, recent posts are checked
for one already linking to the note (a send whose reply was lost).
## Consequences
- Echofeed is no longer needed; it must stop posting to Bluesky or posts
appear twice.
- A post deleted by hand on Bluesky is not re-created.
@@ -0,0 +1,34 @@
# ADR-067: Micropub update, delete, drafts and replies
## Status
Accepted (v2.4.0, 2026-10-31)
## Context
The Micropub endpoint could only create published notes. Apps could not edit
or delete posts, drafts and replies were dropped, and photos sent with the
post or hosted elsewhere were lost.
## Decision
1. **Update** (JSON only, `update` scope). Supported properties are the ones a
note can store: `content`, `name` (the `# Heading` first line),
`category`, `photo`, `in-reply-to`, `post-status`. Anything else is a 400,
not silently ignored. Order: delete, then replace, then add (`add` only for
`category` and `photo`). Success is 204.
2. **Delete is soft** (`delete` scope) so `undelete` can restore the note and
its file from `.trash/`. Crossposts are removed on delete and posted again
on undelete.
3. **Drafts**: `post-status: draft` creates an unpublished note; crossposts
wait until it is published.
4. **Photos**: multipart uploads are saved like media-endpoint uploads; remote
URLs are downloaded only from public IP addresses, up to 20 MB, max 4.
5. **Queries**: `q=source` returns everything create accepts (round-trip);
without `url` it lists recent posts, drafts included, for editing apps.
`q=category` lists tags by use.
## Consequences
- `published` (backdating) can be set on create but not changed by update.
- Likes, reposts and bookmarks remain unsupported (round 2).
@@ -0,0 +1,34 @@
# ADR-068: Likes and reposts
## Status
Accepted (v2.5.0, 2026-11-01). Bookmarks were considered and declined.
> Bookmarks were later added in v2.9.0 (Micropub `bookmark-of`, migration 015) without a separate ADR. (status updated 2026-11-03)
## Decision
1. **Storage.** `notes.like_of` / `notes.repost_of` (migration 013). The note's
content is an optional comment and may be empty.
2. **Display.** Likes stay off the home page and feeds and are listed on
`/likes`. Reposts appear in the stream and feeds. Both show the target's
fetched title and author (the reply-context cache) and carry
`u-like-of` / `u-repost-of`. The target is sent a webmention.
3. **Crossposting, per target.**
| | Target is a post on that network | Anything else |
|---|---|---|
| Like | native favourite / like | nothing |
| Repost, no comment | native boost / repost | "🔁 title url" post |
| Repost, with comment | Bluesky quote post; GoToSocial status | comment + "🔁 title url" post |
GoToSocial finds fediverse posts with `/api/v2/search?resolve=true`
(needs `read:search`; favourites need `write:favourites`). Bluesky
recognises `bsky.app/profile/<actor>/post/<rkey>` URLs. Native actions
are stored as `favourite:<id>` / `reblog:<id>` or the like/repost record
URI, and deleting the note undoes them.
## Consequences
- Editing a like/repost doesn't change a native favourite/boost.
- A repost with a comment can't be a GoToSocial boost (no quote posts there).
@@ -0,0 +1,57 @@
# ADR-069: Crossposting to Threads
## Status
Accepted (v2.6.0, 2026-10-31)
## Context
ADR-065 and ADR-066 crosspost notes to GoToSocial and Bluesky. Threads
(Meta) has a publishing API, so it can be a third target on the same loop.
## Decision
1. **Third client, same loop.** `ThreadsClient` is target id `threads`
behind the `syndications` table and reconcile loop from ADR-065.
2. **Auth.** A long-lived Threads user token (`THREADS_ACCESS_TOKEN`) with
`threads_basic`, `threads_content_publish` and `threads_delete`. These
tokens expire after 60 days, so StarPunk refreshes the token about once a
day (`th_refresh_token`) from the worker loop, even when nothing is being
posted. The current token is kept in `DATA_PATH/threads_token.json`
(mode 0600) with a hash of the configured token. A different configured
token replaces the saved one. A file lock stops gunicorn workers from
refreshing at the same time. The file avoids a migration for one value.
3. **Posting.** Create a container, poll its status until `FINISHED`, then
`threads_publish`. Text-only notes are `TEXT` posts. Threads previews
the first link in the text (the permalink) by itself. An explicit
`link_attachment` is not sent, because Meta rejects one for a page it
hasn't fetched yet ("Invalid Link Attachment", subcode 4279047), which
is always the case for a new note (v2.6.2). One image is an `IMAGE` post and 2–4 images are a
`CAROUSEL`. Meta fetches images from their public `/media/` URL. Only
JPEG and PNG up to 8 MB are allowed, so GIF and WebP images are left off.
Alt text goes in `alt_text`.
4. **Content.** Plain text (as for Bluesky), then the permalink, shortened
to 500 characters. Non-ASCII characters are counted as UTF-8 bytes,
because Threads counts emoji that way. Threads allows one tag per post,
so the note's first tag is sent as `topic_tag` and no hashtags are added
to the text.
5. **Edits don't follow.** The API has no way to edit a post. As with
Bluesky, the post stays as first sent, and unpublishing, unticking the
target or deleting the note removes it.
6. **Likes and reposts (ADR-068).** The API can't like or repost outside
content, so likes are not crossposted. Reposts are text posts with
"🔁 title" and the URL, plus the comment if there is one, and the
reposted page as the link preview.
7. **No double posts.** A retried create first looks for a recent post that
already contains the permalink.
## Consequences
- Meta limits accounts to 250 posts and 100 deletes per day. Rate-limit
errors (Graph codes 4, 17, 32, 613) are retried.
- If StarPunk is down for more than 60 days, the token expires and a new
one is needed.
- Getting the first token requires a Meta developer app with the Threads
use case. See the setup notes in `docs/design/v2.6.0/`.
- Notes with more than five links fail on Threads (a Meta limit). The
error is shown in the editor.
@@ -0,0 +1,44 @@
# ADR-070: Links and link cards in crossposts
## Status
Accepted (v2.6.3, 2026-11-01)
## Context
Every crosspost carries the note's permalink. Notes often link elsewhere
too, usually to the article being discussed. Before this change:
- Bluesky and Threads get plain text, and a Markdown link `[text](url)`
lost its URL, so readers there couldn't follow it.
- The Bluesky card always showed the note, and the Threads preview showed
whichever link came first.
- Threads rejects posts with more than 5 different links.
## Decision
1. **The card shows what the note links to.** If the note has its own link
(a Markdown link, `<url>` or a bare URL), the first one gets the
preview card. A note without links gets a card for itself, as before.
Images still take the place of a card on Bluesky.
2. **Bluesky.** The card is built by StarPunk: it fetches the page's Open
Graph, Twitter or plain HTML title, description and image, and uploads
the image (shrunk to under 1 MB) as the thumbnail. If the page can't
be fetched, the card has just the URL. Markdown link text becomes a
clickable link facet, so no characters are spent on the URL.
3. **Threads.** It has no clickable link text, so Markdown links are
written "text (url)". Threads previews the first link in the text, which
is now the linked page. To stay within Meta's 5-link limit, after 4
different links in the body a Markdown link keeps only its text.
4. **GoToSocial** already gets Markdown, so its links are unchanged.
Receiving servers build their own cards.
5. **Fetching is careful.** Only http(s) URLs on public addresses are
fetched, and this is checked again at every redirect. At most 512 KB of
a page and 5 MB of an image are read. A failed fetch never fails the
post.
## Consequences
- One or two extra requests to the linked site when a note is crossposted
to Bluesky.
- Reposts (ADR-068) are unchanged: their card is the reposted page.
+206
View File
@@ -0,0 +1,206 @@
# ADR-071: WebSub publishing with a built-in hub
## Status
Accepted (v2.10.0, 2026-11-03)
## Context
Readers poll StarPunk's feeds. WebSub (W3C Recommendation,
https://www.w3.org/TR/websub/) lets readers that support it (Monocle,
Aperture, Inoreader, NewsBlur, Miniflux and others) subscribe once and
have new content pushed to them within seconds.
WebSub has three roles. StarPunk is the **publisher**. Someone has to run
the **hub** that keeps subscriptions and delivers content. Public hubs
exist (switchboard.p3k.io, pubsubhubbub.appspot.com), but they are a
third-party dependency and some have shut down. The rest of this site is
self-hosted.
ADR-014 deferred WebSub for v1 as "complexity, external dependency".
Two existing problems matter here:
- Feeds are cached for `FEED_CACHE_SECONDS` (5 minutes) and nothing clears
the cache when a note changes. A hub fetching right after a change would
get stale content. Regular readers are also up to 5 minutes behind.
- The RSS feed is served at `/feed.rss` and `/feed.xml` (legacy alias), and
calls itself `/feed.xml`.
## Decision
StarPunk runs its own hub, **for its own topics only**, at `/websub`.
`WEBSUB_HUB_URL` can point at an external hub instead; then StarPunk
advertises that hub, pings it on changes, and its own `/websub` is off.
### 1. Topics
Absolute URLs under `SITE_URL`:
| Topic | Content type |
|---|---|
| `/` (h-feed) | text/html |
| `/feed.atom` | application/atom+xml |
| `/feed.rss` | application/rss+xml |
| `/feed.json` | application/feed+json |
| `/likes` | text/html |
| `/tag/<name>` | text/html |
`/feed.xml` and the negotiated `/feed` advertise the topic of the format
they serve as `rel=self` (`/feed.rss`, `/feed.atom`, `/feed.json`). The
RSS feed's own self link changes from `/feed.xml` to `/feed.rss`.
### 2. Discovery (publisher, spec 4)
Every topic response carries `Link: <hub>; rel="hub"` and exactly one
`Link: <topic>; rel="self"`. Embedded links as well: `<link rel="hub">` and
`<link rel="self">` in the HTML head of topic pages, `<link rel="hub">`
next to the existing self link in Atom, `<atom:link rel="hub">` in RSS, and
`"hubs": [{"type": "WebSub", "url": ...}]` in JSON Feed.
### 3. What changed, and when
`websub.note_changed(note, old_tags=())` is called wherever
`syndication.note_changed` is, plus on create. It:
1. clears the feed cache (all formats), and
2. if the note is or was public, queues delivery for `/`, the three
feeds, `/likes` if it is a like, and each `/tag/<name>` for its tags
before and after the change.
Drafts that stay drafts change nothing.
### 4. Hub endpoint (spec 5)
`POST /websub`, `application/x-www-form-urlencoded`:
- `hub.mode` subscribe or unsubscribe; `hub.topic` must be one of the
topics above (after normalising), else 400 with a plain-text reason;
`hub.callback` must be an absolute http(s) URL that passes the SSRF
check; `hub.secret` under 200 bytes; `hub.lease_seconds` clamped to
1 day to 30 days (default 10 days). Unknown parameters are ignored.
- Response: 202 Accepted, independent of what verification later finds.
- Limits: the Webmention rate limiter (per IP); at most 25 subscriptions
per callback host and 1,000 in total; beyond that, 429.
### 5. Verification of intent (spec 5.3)
A background worker sends `GET <callback>` with the callback's query
string kept and `hub.mode`, `hub.topic`, `hub.challenge` (random, 32
bytes) and `hub.lease_seconds` appended. Success is a 2xx with the body
equal to the challenge; anything else, including redirects, fails and
leaves the subscription state unchanged. On success a subscribe becomes
active with `expires_at = now + lease`, and an unsubscribe deletes it.
A re-subscribe replaces the secret and lease only once verified.
### 6. Content distribution (spec 6, 7)
For each change, one delivery per active, unexpired subscription to the
topic. The worker renders the topic inside the app (no HTTP round trip
through Caddy), so subscribers get exactly what the public URL serves,
then `POST`s it to the callback with:
- `Content-Type` of the topic,
- `Link` headers with `rel=hub` and `rel=self`,
- `X-Hub-Signature: sha256=<HMAC-SHA256(secret, body)>` if a secret was
given.
A 2xx is success. 410 Gone removes the subscription. Anything else is
retried after 1 min, 5 min, 30 min, 2 h and 6 h, then dropped; a
subscription whose deliveries fail for 3 days is removed. Only the latest
content is sent: a queued delivery for a topic is replaced by a newer one.
### 7. Outbound safety
Callbacks are arbitrary URLs, so verification and delivery go through
`webmention_ssrf` (`safe_get`/`safe_post`): public addresses only, the
connection pinned to the checked address, no redirects, 10 s time limit,
1 MB response cap.
### 8. Data (migration 016)
- `websub_subscriptions(id, topic, callback, secret, lease_seconds,
state ['pending','active'], expires_at, created_at, verified_at,
last_success_at, last_error)`, unique on `(topic, callback)`.
- `websub_intents(id, topic, callback, mode, secret, lease_seconds,
attempts, next_attempt_at, created_at)`: verifications to do.
- `websub_deliveries(id, subscription_id, attempts, next_attempt_at,
last_error, created_at)`, unique on `subscription_id`.
### 9. Worker
The same pattern as the Webmention worker: one thread per gunicorn
process, a file lock so only one process works the queues, app context
around each poll. Expired subscriptions are pruned once a day.
### 10. Admin
The admin dashboard shows active subscribers per topic and failing
deliveries.
### 11. Configuration
`WEBSUB_ENABLED` (default true), `WEBSUB_HUB_URL` (default: built-in),
`WEBSUB_MAX_SUBSCRIPTIONS` (1000), `WEBSUB_MAX_PER_HOST` (25).
## Consequences
- Readers that support WebSub get posts and edits within seconds.
- Feeds are fresh right after a change for every reader, because the cache
is cleared.
- StarPunk sends HTTP requests to subscriber-chosen URLs. The SSRF
protections, per-host and total limits, and rate limiting bound that.
- websub.rocks' hub tests use websub.rocks' own topics, which this hub
refuses by design. Testing uses unit tests of each spec section and a
real subscription from a reader after deploy.
- The RSS self link changes to `/feed.rss`. Readers keep whichever URL
they subscribed to; both keep working.
## Alternatives considered
- **Public hub only.** Least code, but a third-party dependency for a
self-hosted site. Still available through `WEBSUB_HUB_URL`.
- **A separate hub service** (another container). More to deploy and
keep updated, and few maintained hubs exist.
- **Accepting any topic** (a general-purpose hub). Not needed, and it
would let the hub be used to fetch and push arbitrary URLs.
## Amendments (v2.10.1, 2026-11-03)
An adversarial review of v2.10.0 against the Recommendation led to these
changes; where they differ, they replace the text above.
- **Worker robustness.** Each queued item is handled on its own: an
unexpected error drops (verification) or retries (delivery) that item
only. Before, a callback answering with a large page or containing a
control character stopped the worker for good.
- **Time limits.** `safe_get` and `safe_post` close the connection when
the whole exchange exceeds its time limit (10 s), not just a single read.
Verification replies over 4 KB fail; delivery responses are read up to
4 KB.
- **Leases are kept (spec 7).** A failing subscriber is not removed; its
deliveries are retried and then given up, and the next change is
delivered again. Only lease expiry, 410 Gone or unsubscribing end a
subscription. DNS failures are retried. A topic that renders 404/410
ends its subscriptions with a `hub.mode=denied` notice; other render
errors are retried.
- **Photos.** A note with photos is announced once they are attached
(`attach_media_to_note` notifies), and the feed cache key and ETag
include `content_version`, so photo and caption changes reach feeds and
subscribers. ETags no longer depend on the feed's build time.
- **Subscription requests (spec 5.1).** Unreserved percent-encodings are
decoded; the Content-Type check is case-insensitive; `hub.secret` and
`hub.lease_seconds` are ignored for unsubscriptions; an empty secret is
refused; verification echoes `hub.topic` exactly as requested
(`requested_topic`); unsubscribing works while an external hub is
configured. Callbacks must be printable, at most 2,048 characters, not on
this site, on port 80, 443 or 1024+, and are stored with an IDNA host.
- **Limits** count pending requests as well as subscriptions, per callback
*address* (so subdomains on one server share a limit), with at most 200
pending requests. Rate limiting needs `TRUSTED_PROXY=true` behind Caddy.
- **Races.** Each request has a `request_id`; verifying an older request
doesn't consume a newer one (migration 017).
- **Less work.** A topic is rendered once per worker pass; likes only
update `/likes` and tag pages.
- Corrections: there is no `state` column (a subscription row exists only
once verified); expired leases are pruned hourly.
@@ -0,0 +1,51 @@
# ADR-072: Remove the metrics dashboard
## Status
Accepted (v2.12.0, 2026-11-03). Supersedes ADR-053 and ADR-060.
## Context
v1.1.1 and v1.1.2 added built-in performance monitoring (ADR-053): HTTP
request timing middleware, a wrapper on every database connection that
timed queries, a memory-sampling thread, feed-generation counters, an
in-memory buffer of the results, and an HTML dashboard at
`/admin/metrics-dashboard` with a JSON twin at `/admin/metrics` (ADR-060
covered adapting the data for the dashboard template).
In practice it told us little:
- The buffer lives in memory in each of the four gunicorn processes, so the
dashboard showed whichever process served the request, and everything was
lost on restart or deploy.
- Nothing in it was ever used to find or fix a problem. Logs (with
correlation IDs), `/health` and `/admin/health` were.
- It cost something on every request and every query, kept a thread per
process, and needed `psutil`, about 1,300 lines of code and three test
files.
## Decision
Remove the dashboard and everything that only fed it: the `starpunk/monitoring`
package, the HTTP metrics middleware (including its `X-Request-ID`
response header), `MonitoredConnection`, the memory monitor, the
feed-generation tracking calls, `/admin/metrics-dashboard`,
`/admin/metrics`, the `METRICS_*` settings and the `psutil` dependency.
Keep:
- `/health` and `/health?detailed=true`;
- `/admin/health`, without its `metrics` section (database, pool
statistics, filesystem and disk checks stay);
- connection pool statistics (`get_pool_stats`) and the feed cache, which
do work of their own;
- structured logs with correlation IDs (ADR-054).
## Consequences
- Less code and one fewer dependency; no timing work on each request or
query; one fewer thread per process.
- Slow database queries are no longer logged. If that is ever needed, a
query timer can be added at the pool level, or an external tool used.
- `METRICS_*` environment variables are ignored if still set.
- Bookmarks to `/admin/metrics-dashboard` now get a 404.
+281
View File
@@ -0,0 +1,281 @@
# ADR-073: Backfeed of responses to crossposts
## Status
Accepted (v2.13.0, 2026-11-10; amended v2.13.2, v2.13.3)
Implementation notes (v2.13.0):
- Until an avatar has a local copy, pages show the initial and keep the
original URL only as `<data class="u-photo">` for microformats parsers,
so browsers don't load it.
- Bluesky post views carry no bio, so reply and quote authors' profiles
are fetched (`getProfiles`, 25 at a time) to honour `#nobridge`.
- A Threads token without `threads_manage_insights` stops Threads
backfeed as a whole, as for any refused credential.
- Backfill checks take a shared slot in `site_state`
(`backfeed_backfill_last`), so all gunicorn workers together do at most
one every 5 seconds.
## Context
StarPunk crossposts notes to GoToSocial, Bluesky and Threads (ADR-065,
ADR-066, ADR-069). People like, repost, quote and reply to those copies,
and none of that reaches the note. [Backfeed](https://indieweb.org/backfeed)
brings those responses back to the original post.
The usual tool is [Bridgy](https://brid.gy/about), which polls the silos and
sends webmentions. It is a poor fit here:
- It supports Mastodon and Bluesky, not Threads.
- It only checks the 30 most recent posts, about every 30 minutes, and
says nothing when a like is withdrawn or a reply deleted.
- Since v2.8.0 StarPunk rejects webmentions from a note's own crossposts
(`_is_own_syndicated_copy`, `webmention_worker.py`). Bridgy's like and
repost pages most likely carry the crosspost's URL plus a fragment as
`u-url`, so they would be rejected. Replies have their own URL and get
through (production webmention #1 was a Bridgy reply from Mastodon).
- Every Bridgy source is on `brid.gy`, which shares one source-domain rate
limit of 10 webmentions a minute (ADR-063, Decision 10).
- Every response passes through a third party.
StarPunk already holds a credential for each silo and knows each copy
(`syndications.remote_id`, `remote_url`). What the silos let it read:
| Silo | Likes | Reposts | Quotes | Replies | Credential change |
|---|---|---|---|---|---|
| Bluesky | `getLikes` | `getRepostedBy` | `getQuotes` | `getPostThread` | none (app password) |
| GoToSocial | `favourited_by` | `reblogged_by` | not supported | `context` | read scopes |
| Threads | count only | count only | count only | `replies` | `threads_read_replies`, `threads_manage_insights` |
The Threads API never says who liked, reposted or quoted a post; the
Insights API gives lifetime counts only.
The production GoToSocial prunes remote posts older than 90 days, but never
a thread that a local account started or took part in (GoToSocial's "Post
Caching and Pruning" docs), so replies to StarPunk's copies are kept. On
2026-11-10 the production database still had 106 remote replies to the
account older than 90 days, back to August 2024.
## Decision
### 1. Native polling, no Bridgy
StarPunk polls each silo itself for responses to its own copies. Bridgy is
switched off for the Mastodon and Bluesky accounts once this ships. The
own-copy rule stays as it is.
### 2. Backfed responses are webmentions
Responses are stored in the `webmentions` table, so moderation, the
Responses section, and re-sending webmentions when a note's responses
change (spec 3.1.4, `note_responses_changed`) work as they do now. A
backfed row:
- `source_url`: the response's own URL where it has one (a reply or quote
post). Likes and reposts have no page, so they get the copy's URL plus
`#liked-by-<actor id>` or `#reposted-by-<actor id>`. The
`(source_url, target_url)` key makes repeated checks update the same row.
- `target_url`: the note's permalink. `target_note_id`: the note.
- `origin`: new column. `webmention` for received webmentions (the
default), else `gotosocial`, `bluesky` or `threads`.
- `remote_id`: new column. The silo's id for the response (status id,
AT-URI, Threads media id), or the actor's id for a like or repost.
- `mention_type`: `like`, `repost`, `reply` or the new `quote`.
- Author name, profile URL and avatar from the silo's account data;
`content_snippet` from the reply or quote text, sanitised the same way as
received webmentions (Mastodon HTML through nh3, Bluesky and Threads text
escaped with links kept); `source_published` from the silo's timestamp.
- `validation_status = 'valid'`: StarPunk read it from the silo's API
itself, so there's nothing to fetch and verify. Moderation follows
`WEBMENTION_AUTO_APPROVE` like any first-time webmention.
Only replies directly to the copy are taken, not deeper replies in the
thread. Responses from the accounts StarPunk posts as are skipped.
### 3. Schedule
Each synced copy of a published note is checked on a schedule that slows
as the note ages:
| Note age | Check every |
|---|---|
| under 1 day | 15 minutes |
| 1 to 7 days | 1 hour |
| 7 to 30 days | 1 day |
| 30 to 90 days | 1 week |
| over 90 days | not checked |
A random jitter of up to 10% spreads the checks. The cutoff is
`BACKFEED_MAX_AGE_DAYS` (default 90). Copies of likes and reposts
(ADR-068) aren't checked: they have no responses of their own. A note that
is unpublished or deleted is not checked, and its rows stay as they are.
### 4. Withdrawn likes and deleted replies
A check that read the whole list (every page, under a cap of 1,000 items
per list) marks that note's rows from that silo that are no longer there
as `validation_status = 'invalid'` with the error "No longer on
<silo>". They drop off the page but keep their moderation decision, and
come back if they reappear. A check that stopped early (an error, or the
cap) removes nothing.
### 5. Privacy
- Only public responses: GoToSocial `visibility` `public` (not unlisted,
followers-only or direct); on Bluesky, nothing from accounts with the
`!no-unauthenticated` label.
- People who opt out are skipped: `#nobridge` or `#nobot` in their bio
(Mastodon `note`, Bluesky `description`), the same signal Bridgy honours.
A row already stored for them is removed at the next check.
- Admin **Delete** stops being a real `DELETE` and sets
`moderation_status = 'deleted'` (the value exists but is unused), so a
later check or re-sent webmention can't bring the response back. This
applies to all webmentions.
### 6. Avatars are cached locally
Received webmentions hotlink author photos today. That leaks visitors'
addresses to other sites, and Threads and Bluesky avatar URLs change or
expire. Avatars are fetched once with `safe_get` (1 MB limit), resized to
96 px WebP with Pillow and saved as `DATA_PATH/avatars/<sha256 of URL>.webp`,
served from `/avatars/`. They're refetched when the URL changes, and files
no row uses are removed weekly. Received webmentions use the same cache
from their validation step. A failed fetch falls back to the coloured
initial.
### 7. Threads counts
Threads likes, reposts and quotes are shown as anonymous counts, not as
responses. The worker reads
`/{media-id}/insights?metric=likes,reposts,quotes` with each Threads check
and stores the numbers on the copy's `syndications` row. The note shows a
plain line such as "On Threads: 12 likes · 3 reposts" after the
Responses, with no `u-like` or `h-cite` markup, because these aren't
people. Threads `replies` counts aren't used: the replies themselves are
shown. `views` and `shares` are left out (Meta marks them "in
development").
### 8. Backfill
An admin action, **Backfill responses**, checks every synced copy once,
whatever the note's age. Nothing runs by itself on deploy; the owner runs
it after the first deploy (and may run it again later). It:
- marks every synced copy due now, flagged as a backfill;
- is worked through by the same worker, at most one copy every 5 seconds,
ahead of the routine schedule only when nothing routine is due;
- records when it started and finished; the admin page shows this, and
the button is disabled while one is running.
Upserts make it safe to repeat.
### 9. Worker
The checks run in the syndication worker's loop (ADR-065) after crossposts,
reusing its clients, token refresh (`maintain()`) and the lease pattern:
a copy is claimed with `backfeed_claimed_until` so only one gunicorn worker
checks it. A failed check backs off like a failed crosspost
(`min(30 s × 2^attempts, 6 h)`) without changing the copy's crosspost
state. If a silo answers with a permission error, backfeed for that silo
stops until restart and the admin page says which credential needs which
scope; crossposting is unaffected.
Per check: Bluesky 4 calls (likes, reposts, quotes, thread), GoToSocial 3,
Threads 2 (replies, insights), plus extra pages.
### 10. Data (migration 020)
- `webmentions`: rebuilt to add `quote` to the `mention_type` CHECK, with
new columns `origin TEXT NOT NULL DEFAULT 'webmention'` and
`remote_id TEXT`, and an index on `(target_note_id, origin)`.
- `syndications`: new columns `backfeed_next_at`, `backfeed_checked_at`,
`backfeed_claimed_until`, `backfeed_attempts`, `backfeed_error`,
`backfeed_backfill` (flag), and `likes_count`, `reposts_count`,
`quotes_count` (Threads only).
- The backfill's start and finish times go in the existing `site_state`
key/value table (migration 016): `backfeed_backfill_started`,
`backfeed_backfill_finished`.
### 11. Display and admin
- Each backfed response shows a small "via Bluesky" (or GoToSocial,
Threads) label. Quotes get their own group, rendered like mentions with
the quote text, as `u-mention h-cite`.
- The note editor shows when responses were last checked, with a
**Check now** button for each copy.
- The webmentions list can be filtered by origin.
- The dashboard shows backfeed errors next to crosspost errors.
### 12. Configuration and credentials
- `BACKFEED_ENABLED` (default `false`; production sets `true`) and
`BACKFEED_MAX_AGE_DAYS` (default 90).
- GoToSocial: a new token with read access added to the current scopes
(`read:statuses` and `read:accounts`, or plain `read`; checked against
GoToSocial's scope list when the token is issued).
- Threads: the Meta app is re-authorised with `threads_read_replies` and
`threads_manage_insights` added. The new `THREADS_ACCESS_TOKEN` replaces
the saved token file automatically (ADR-069, Decision 2).
- Bluesky: no change.
- Production tokens go in the Ansible vault (`bin/vault-run`).
While at it, the Threads client moves from `graph.threads.net` to
`graph.threads.com`, the host Meta's current documentation uses.
## Amendments (v2.13.2, 2026-11-10)
- **Display.** Likes, reposts and bookmarks are a line of counts under the
note instead of facepiles in the Responses section. Hovering or focusing
a count lists the people (as `u-like` / `u-repost` / `u-bookmark`
h-cites) and the silo each came from. Threads' anonymous counts are added
to the totals ("+N on Threads") instead of a separate line (Decision 7).
- **Web actions.** Each like, repost and reply count is an
`<indie-action do="…" with="<permalink>">` whose fallback is a link to
the same action on each crosspost. No web-action polyfill is loaded;
handlers such as browser extensions can use the markup.
## Amendments (v2.13.3, 2026-11-10)
- **Opted-out people are counted.** Their likes, reposts and quotes are
still not stored or named (Decision 5), but each check records how many
there were on the copy's `syndications` row (`likes_count`,
`reposts_count`, `quotes_count`, the columns Threads' counts use), and
the counts line adds them to the totals as "N on <silo>". Totals then
match the silo.
## Consequences
- Responses on all three silos show on the note within 15 minutes for new
notes, with no third party involved.
- Withdrawn likes and deleted replies disappear by themselves for 90 days;
after that a note's responses are frozen.
- Threads shows replies in full but likes, reposts and quotes only as
numbers.
- Backfeed stores other people's public posts and avatars. The privacy
rules (Decision 5) and moderation are the controls; there is no consent
step.
- More API calls. At about 50 new notes a month the routine schedule is
roughly 5,000 calls a month across the three silos, well inside their
limits. The backfill makes one burst of a few calls per copy.
- Deleting a webmention in the admin now leaves a tombstone row.
- A crosspost removed and posted again gets a new copy; responses to the
old copy are marked gone at the next check.
## Alternatives considered
- **Bridgy.** No Threads, 30-post window, no withdrawals, a third party in
the path, and it needs the own-copy rule and rate limit loosened (see
Context). Rejected.
- **A separate responses table.** Would duplicate the display, moderation
and 3.1.4 re-sending code that `webmentions` already has. Rejected.
- **Notification feeds** (Mastodon `/api/v1/notifications`, Bluesky
`listNotifications`): one call covers every post, but they only report
new events, so withdrawals are never seen and a backfill can't use them.
Could be added later as a fast path in front of per-post checks.
- **Streaming** (Bluesky Jetstream, Mastodon streaming API): real time, but
needs a long-lived connection per gunicorn worker. Not worth it for a
personal site.
- **Keep the Threads counts out.** They aren't IndieWeb responses, but they
are the only signal Threads gives, and showing them anonymously costs one
call and identifies no one. Kept.
+155
View File
@@ -0,0 +1,155 @@
# ADR-074: Notifications through Apprise
## Status
Accepted (v2.14.0, 2026-11-10)
## Context
Responses now reach a note from two places: received webmentions
(ADR-063) and backfeed from the crossposts (ADR-073). The owner only finds
out about them by opening the site or the admin. Crossposts that finally
fail, and silos that refuse a backfeed credential, are just as invisible.
Owners use different notification services. The production site uses
Gotify (read on iOS through iGotify). Others use ntfy, Pushover, email,
Matrix, Discord and so on. [Apprise](https://github.com/caronc/apprise)
covers about 100 of them behind one URL syntax (`gotifys://host/token`,
`ntfys://topic`, `mailtos://…`), both as a Python library and as a
separate HTTP server (apprise-api).
Options considered:
1. **Gotify only.** Simplest for this site, no use to anyone else.
2. **Post to an apprise-api server.** Standard, but every owner needs one
running, and messages take an extra hop.
3. **Embed the Apprise library.** Sends straight to each service; an owner
who prefers a central apprise-api can still use its `apprise://` /
`apprises://` URL.
4. **Gotify plus a generic JSON webhook, hand-written.** No dependency, but
every other service needs glue the owner has to write.
## Decision
### 1. Embed Apprise
StarPunk depends on the `apprise` package and sends with it directly.
`NOTIFY_URLS` holds one or more Apprise URLs (space- or comma-separated);
each message goes to all of them. With no URLs, notifications are off and
nothing else changes.
### 2. What can be notified
`NOTIFY_EVENTS` is a comma-separated list. Default:
`reply,mention,quote,problems`.
| Event | Sent when |
|---|---|
| `reply`, `mention`, `quote`, `like`, `repost`, `bookmark` | a response of that type is first shown on a note, or first awaits moderation (received webmention or backfeed alike) |
| `problems` | a crosspost fails for good (permanent error or last attempt), or a silo refuses the backfeed credential (ADR-073 Decision 9) |
### 3. Messages
Markdown, so services that render it (Gotify, iGotify, ntfy, Matrix…)
show links; others get the text.
- **Replies, mentions, quotes:** one message each. Title "↩ Reply from Ana
Lima via Bluesky" (or "💬 Mention from …", "❝ Quote from …"); body the
response text (plain, up to 300 characters), then a link to the note's
responses (`<permalink>#responses`).
- **Likes, reposts, bookmarks:** grouped per note and type in each pass:
"♥ 3 new likes on 'Starpunk now supports…'", with up to five names and a
link to the note.
- **Awaiting moderation:** when `WEBMENTION_AUTO_APPROVE` is off, the title
says so and the link goes to the admin webmention page instead.
- **Problems:** "⚠ Crosspost to Bluesky failed" / "⚠ Backfeed for Threads
stopped" with the error and a link to the note's edit page or the
Webmentions page.
Apprise's Gotify plugin has no click-through link, so the link is in the
text.
### 3a. Priority
Every message is sent with the same priority, `NOTIFY_PRIORITY` (default
`normal`). Apprise takes priority as a `priority` argument in each service
URL, with values that differ by service, so StarPunk adds
`priority=<NOTIFY_PRIORITY>` to each URL in `NOTIFY_URLS` that doesn't set
one itself; a URL can still set its own. Apprise's names (`low`,
`moderate`, `normal`, `high`, `emergency`) work for most services; Gotify
also takes 0–10 (it maps the names to 0, 3, 5, 8, 10). Different
priorities per event can come later.
### 4. Delivery
- New column `webmentions.notified_at`. A sweep finds responses with
`notified_at IS NULL`, `validation_status = 'valid'`, a moderation status
of `approved` or `pending`, and a type in `NOTIFY_EVENTS`, sends them,
and sets `notified_at`. Types not selected are marked without sending.
- The sweep runs in the webmention worker loop, which already runs in one
process at a time (its file lock), so nothing is sent twice by several
gunicorn workers. It handles received and backfed responses alike.
- A failed send leaves `notified_at` empty, so the next pass (about 10
seconds later) retries, with backoff up to 15 minutes. Responses older
than 24 hours that still haven't gone out are marked and dropped, with a
warning, so an outage doesn't end in a flood. The last error is shown on
the admin Webmentions page.
- Problems are sent at the moment they happen (from the syndication and
backfeed workers) and not retried: they are also in the log and admin.
- Sending never blocks or breaks validation, crossposting or backfeed.
### 5. No floods from history
- Migration 021 sets `notified_at` on every existing response.
- Responses stored by a backfill check (ADR-073 Decision 8) are stored with
`notified_at` set.
- So are responses found on a crosspost's first check when its note is more
than a day old (copies linked after the fact).
- Grouping (Decision 3) keeps a busy note to one like message per pass.
### 6. Admin
- **Send test notification** on the admin Webmentions page sends a test
message to every URL and reports each result.
- The page shows whether notifications are on, which events are selected,
and the last send error.
### 7. Configuration
| Setting | Default | Meaning |
|---|---|---|
| `NOTIFY_URLS` | empty (off) | Apprise URLs, space- or comma-separated |
| `NOTIFY_EVENTS` | `reply,mention,quote,problems` | which events are sent |
| `NOTIFY_PRIORITY` | `normal` | priority of every message, added to URLs that don't set one |
Settings stay in the environment, like the rest of StarPunk's (ADR-052).
URLs contain tokens and belong with the other secrets.
## Consequences
- One new dependency, `apprise`, with its own (requests, PyYAML, click,
Markdown, certifi). It must install on Python 3.14 in the container
image; this is checked before the release.
- Owners can send to almost any service without writing glue, directly or
through apprise-api.
- Each response is notified at most once, a few seconds after it appears,
and only if its type is selected.
- Apprise opens outbound connections to whatever URLs the owner configures,
outside the SSRF-safe fetcher. That is intended: the owner chooses these
URLs, as with the crosspost targets. A service behind the same reverse
proxy as StarPunk may need the deployment's internal-address entry
(`extra_hosts`), as GoToSocial does.
- Notification text includes other people's names and words, sent to the
owner's own services only.
## Alternatives considered
- **Gotify only / hand-written webhook.** See Context, options 1 and 4: no
dependency, but each service beyond Gotify needs glue the owner writes.
- **apprise-api only.** See option 2. Still possible through an
`apprises://` URL.
- **Sending from where responses are stored** (webmention worker,
backfeed) instead of a sweep. Two code paths, no retry, and backfeed
runs in every gunicorn worker. Rejected.
- **Choosing events in the admin UI.** Possible later; the environment is
where StarPunk's settings live today.
+121
View File
@@ -0,0 +1,121 @@
# ADR-075: MCP server for AI clients
## Status
Accepted (v2.15.0, 2026-11-11)
## Context
The owner wants AI assistants (claude.ai custom connectors, Claude Code and
other [Model Context Protocol](https://modelcontextprotocol.io/) clients) to
read, search and write notes. Micropub already covers writing, but MCP
clients don't speak Micropub: they discover tools over MCP and get tokens
through OAuth 2.1 with protected resource metadata (RFC 9728), resource
indicators (RFC 8707), and either dynamic client registration or a Client ID
Metadata Document (CIMD).
StarPunk already runs an OAuth/IndieAuth authorization server (ADR-064) with
PKCE, public clients only, and clients identified by a `client_id` URL whose
JSON document lists `redirect_uris`. That is what MCP calls CIMD. Claude uses
CIMD when the server's metadata says `client_id_metadata_document_supported:
true` and lists `none` in `token_endpoint_auth_methods_supported`.
StarPunk is Flask on gunicorn sync workers (WSGI). The official MCP Python
SDK is ASGI.
Options considered:
1. **Local stdio bridge** that calls `/micropub` with a stored token, like
existing Micropub MCP servers. No server changes, but every client needs
the bridge installed, claude.ai and mobile can't use it, and reading is
limited to what Micropub queries expose.
2. **Hand-written MCP route in Flask**, stateless JSON only. No new
dependencies or processes, but StarPunk would own the protocol details
(version negotiation, JSON-RPC, Origin checks, auth metadata) and track
the spec itself.
3. **Official SDK in a sidecar process** from the same image, behind the
same reverse proxy, sharing the data volume.
4. **Dynamic client registration** (RFC 7591) in the authorization server.
Every connection would register a new client, and nothing needs it once
CIMD works.
## Decision
### 1. A sidecar running the official SDK (option 3)
`starpunk/mcp_server/` builds an `mcp.server.MCPServer` (SDK 2.3) and serves
it with uvicorn (`uvicorn starpunk.mcp_server.app:app --port 8001`), as a
second container from the same image. The reverse proxy sends `/mcp*` and
`/.well-known/oauth-protected-resource*` to it and everything else to
gunicorn. The endpoint is `SITE_URL/mcp`.
The transport is stateless Streamable HTTP with plain JSON responses. No
session lives in the process, so restarts and several workers are harmless.
The sidecar builds the Flask app with `BACKGROUND_WORKERS_ENABLED=false`, so
only the main container runs the webmention, WebSub and syndication
workers. Each tool runs in a worker thread inside a Flask request context
for `SITE_URL`, and calls the same service functions as the web app.
SQLite (WAL) handles the second process like another gunicorn worker.
### 2. StarPunk is the authorization server; the sidecar only checks tokens
- The authorization server metadata adds
`client_id_metadata_document_supported: true` and
`token_endpoint_auth_methods_supported: ["none"]`, and `scopes_supported`
now includes `draft`.
- An authorization request may carry `resource` (RFC 8707). The only resource
StarPunk names is `SITE_URL/mcp`; anything else is `invalid_target`. The
resource is stored on the code and the token (migration 022), shown on the
consent screen, and must match if the token request repeats it.
- An MCP request with no `scope` is offered `create update delete media
draft`. The owner unticks what they don't want on the consent screen.
(The SDK publishes the resource's `scopes_supported` from its
required scopes, which StarPunk leaves empty so any token can connect.)
- The sidecar's `TokenVerifier` looks tokens up with `verify_access_token`
and reports their resource. With `validate_token_resource`, the SDK
refuses tokens not issued for `/mcp`, so a Micropub app's token can't be
replayed against MCP. MCP tokens still work at `/micropub`; both stay the
owner's own clients, and the Connected Apps page marks MCP tokens.
- DNS rebinding protection allows only the `SITE_URL` host and origin
(plus localhost in `DEV_MODE`).
### 3. Tools: reading and writing notes, nothing else
| Tool | Scope |
|---|---|
| `search_notes`, `list_notes`, `get_note`, `get_site_config` | any token; drafts need `create`, `update`, `delete` or `draft` |
| `list_tags` | `create`, `update`, `delete` or `draft` (as Micropub `q=category`) |
| `create_note` | `create` or `draft` (draft-only tokens always make drafts) |
| `update_note` | `update` |
| `delete_note`, `undelete_note` | `delete` |
| `upload_media` | `media` or `create` |
Writing tools build a Micropub JSON request and call `handle_create`,
`handle_update` and `handle_delete`. Scope rules, validation, crossposting,
webmentions and WebSub notifications are therefore identical to Micropub.
`create_note` makes a draft unless told otherwise. Tool annotations mark
reading tools read-only and update and delete as destructive.
Syndication management, webmention moderation and other admin functions are
not exposed.
## Consequences
- AI clients connect with just the URL: claude.ai via Settings → Connectors,
Claude Code via `claude mcp add --transport http starpunk SITE_URL/mcp`.
The owner approves them on the usual consent screen, and they appear (and
can be revoked) under Connected Apps.
- New runtime dependencies: `mcp` (with Starlette, pydantic and anyio) and
`uvicorn`. They are only imported by the sidecar.
- Deployments run a second container and need two proxy routes. A site that
doesn't run the sidecar just has no MCP; nothing else changes.
- There are still no refresh tokens. MCP clients re-authorize after
`INDIEAUTH_TOKEN_LIFETIME_DAYS`.
- SDK upgrades need care: 3.0 is expected to change defaults (for example
`validate_token_resource`, which StarPunk already sets).
## Alternatives
Options 1, 2 and 4 above. A stdio bridge could still be added later for
clients that can't do OAuth; it would use a token from this flow.
@@ -0,0 +1,74 @@
# ADR-076: Article titles and summaries stored apart from the text
## Status
Accepted (v2.16.0, 2026-11-11)
Article URLs (`/article/<slug>`) and title slugs: [ADR-077](ADR-077-article-urls-and-title-slugs.md) (v2.17.0).
Supersedes the v2.3.3 design ([micropub-title-and-source-url](../design/v2.3.3/2026-10-31-micropub-title-and-source-url.md))
and Q22's "a note has a title when its content starts with `#`". Amends
ADR-014 (feed item titles) and ADR-040 (`p-name` and `p-summary`).
## Context
An IndieWeb article is an h-entry with a `name` that isn't just the start of
its content (W3C Post Type Discovery). Since v1.2.0, StarPunk marked an article
by a `# Heading` first line in the Markdown, and v2.3.3 stored a Micropub `name`
as that heading. A review against IndieWeb practice found four gaps:
1. Feeds repeated an article's title: the item title, then the same heading
rendered at the top of the content.
2. Every feed item had a title, made from a note's first line (or the date),
so feed readers showed notes as headlines and couldn't tell them from
articles.
3. Micropub `summary` was used only when `content` was missing; sent
together with content, it was dropped.
4. Any note that happened to start with a heading became an article, and its
`q=source` changed from what the client sent. Micropub and mf2 treat `name`
and `content` as separate properties.
## Decision
- **Columns.** `notes.title` and `notes.summary` (migration 023). A note with
a title is an article; its Markdown file holds only the body. A heading in
the body is just a heading.
- **Moving existing titles.** Migration 023 has a Python data step
(`migrations.DATA_MIGRATIONS`) that runs once, in the migration's
transaction: for each note whose file starts with a heading (`split_title`,
trashed files included), it moves the heading into `title`, rewrites the file
and updates `content_hash`, then empties `notes_fts` so startup rebuilds it.
If anything fails, the changed files are written back and the transaction is
rolled back, so 023 is retried on the next start. Tying it to the migration
matters: run on every start it would turn new notes' headings into titles.
- **Micropub.** `name` → `title`; `summary` → `summary` when `content` is also
sent. A summary on its own is still stored as the content, which is how Post
Type Discovery reads it. `summary` can be replaced and deleted; `q=source`
returns `name` and `summary` when set. If the content starts with a heading
equal to `name`, that heading is dropped so it isn't shown twice.
- **Feeds.** Only articles have a title: RSS leaves `<title>` out (an item
needs a title or a description, and every item has a description), JSON Feed
leaves `title` out, and Atom, which requires `<title>`, has an empty one.
Atom `<summary>` and JSON Feed `summary` carry the author's summary. Content
no longer contains the title, so it isn't repeated.
- **Pages.** The article page has `p-name`, `p-summary` (when set) and the body
as `e-content`. On the homepage and tag pages an article shows its title,
summary (or the start of its text) and "Read more", without images. The
photos stay in the markup as hidden `u-photo` values so parsers read the same
post type there as on the article page.
- **Crossposting.** Articles go to GoToSocial, Bluesky and Threads as title,
summary (or the start of the text), link and hashtags, without images, the
usual POSSE form for articles. On Bluesky the link card is the article
itself. Notes are posted in full as before.
- **Editor.** Title and Summary fields; a titled note may have no body.
## Consequences
- The migration rewrites note files. Back up the database *and* `data/notes`
(and `.trash`) before deploying.
- `Note.title` is `None` for notes. Places that need a name for any post use
`Note.label` (title, else excerpt, else slug): notifications and the
Bluesky card. `Note.has_title` and `Note.body_html` are gone.
- Search indexes the title column; untitled notes are still listed by their
first line.
- A crossposted article that is edited is re-posted as a teaser.
@@ -0,0 +1,46 @@
# ADR-077: Article URLs and slugs from titles
## Status
Accepted (v2.17.0, 2026-11-12)
Amends ADR-062 (default slugs) for articles and ADR-076 (articles).
## Context
ADR-076 (v2.16.0) made a note with a title an article, but articles still got
a `/note/<timestamp>` URL like any other note. An article is not a note, and
its URL should say so. ADR-062 chose timestamp slugs because short notes have
no good words for a URL and their text may be private; an article has a title
written to be read, and that is the usual source of its slug on the IndieWeb.
## Decision
- **URLs.** Articles live at `/article/<slug>`, other notes at `/note/<slug>`.
`Note.permalink`, `Note.url`, `Note.endpoint` and `utils.post_path` /
`post_url` build them; nothing else writes `/note/` by hand. Slugs stay
unique across both, so a slug still names one post.
- **One URL each, the other redirects.** `/note/<slug>` for an article and
`/article/<slug>` for a note answer 301 to the right one. Adding or removing
a title changes the URL, and links made before (webmentions sent, crossposts,
bookmarks) keep working. A deleted post is 410 at either.
- **Slug from the title.** When a post is created with a title and no
`mp-slug`, the slug is the title made URL-safe (`sanitize_slug`: accents
folded, lowercase, hyphens), cut at a word boundary to 80 characters, `-2`,
`-3`… if taken, `-article` if reserved. A title with nothing usable (only
emoji, say) falls back to the timestamp slug. Notes keep timestamp slugs.
- **Slugs don't follow later title changes.** Editing an article's title, or
making a note an article, keeps its slug: a slug is the stable part of the
URL.
- **Micropub.** `Location` and `q=source` give the new URLs. Updates accept
either URL. An update that adds or removes a title answers 201 with the new
URL in `Location`, as the Micropub spec requires when a post's URL changes;
other updates are still 204.
- **Incoming webmentions** to either URL find the post.
## Consequences
- Production had no articles when this shipped, so no published URL changed.
- A note given a title later keeps its timestamp slug under `/article/`.
- Stored URLs (webmention sources, crosspost records) are not rewritten; the
redirects cover them.
+169 -99
View File
@@ -1,139 +1,209 @@
# Architectural Decision Records (ADRs) Index
This directory contains all Architectural Decision Records for StarPunk CMS. ADRs document significant architectural decisions, their context, rationale, and consequences.
This directory holds StarPunk's Architectural Decision Records. Each ADR
records one decision: its context, what was decided, and the consequences.
Superseded ADRs are kept as history; their Status line names the ADR that
replaced them.
## ADR Format
**Total ADR files**: 73 (72 numbers: ADR-001 to ADR-044 and ADR-050 to
ADR-077, with two files numbered ADR-012). See [Numbering](#numbering).
Each ADR follows this structure:
- **Title**: ADR-NNN-brief-descriptive-title.md
- **Status**: Proposed, Accepted, Deprecated, Superseded
- **Context**: Why we're making this decision
- **Decision**: What we decided to do
- **Consequences**: Impact of this decision
Status values: **Accepted** (in force), **Superseded** (replaced by a later
ADR), **Deprecated** (withdrawn, nothing replaces it; none at present),
**Proposed** (not implemented). "Accepted*" means accepted, with later
changes noted in the ADR's Status section.
## All ADRs (Chronological)
## Current decisions by area
### Foundation & Technology Stack (ADR-001 to ADR-009)
- **[ADR-001](ADR-001-python-web-framework.md)** - Python Web Framework Selection
- **[ADR-002](ADR-002-flask-extensions.md)** - Flask Extensions Strategy
- **[ADR-003](ADR-003-frontend-technology.md)** - Frontend Technology Stack
- **[ADR-004](ADR-004-file-based-note-storage.md)** - File-Based Note Storage
- **[ADR-005](ADR-005-indielogin-authentication.md)** - IndieLogin Authentication
- **[ADR-006](ADR-006-python-virtual-environment-uv.md)** - Python Virtual Environment with uv
- **[ADR-007](ADR-007-slug-generation-algorithm.md)** - Slug Generation Algorithm
- **[ADR-008](ADR-008-versioning-strategy.md)** - Versioning Strategy
- **[ADR-009](ADR-009-git-branching-strategy.md)** - Git Branching Strategy
### Foundation and process
### Authentication & Authorization (ADR-010 to ADR-027)
- **[ADR-010](ADR-010-authentication-module-design.md)** - Authentication Module Design
- **[ADR-011](ADR-011-development-authentication-mechanism.md)** - Development Authentication Mechanism
- **[ADR-016](ADR-016-indieauth-client-discovery.md)** - IndieAuth Client Discovery
- **[ADR-017](ADR-017-oauth-client-metadata-document.md)** - OAuth Client Metadata Document
- **[ADR-018](ADR-018-indieauth-detailed-logging.md)** - IndieAuth Detailed Logging
- **[ADR-019](ADR-019-indieauth-correct-implementation.md)** - IndieAuth Correct Implementation
- **[ADR-021](ADR-021-indieauth-provider-strategy.md)** - IndieAuth Provider Strategy
- **[ADR-022](ADR-022-auth-route-prefix-fix.md)** - Auth Route Prefix Fix
- **[ADR-023](ADR-023-indieauth-client-identification.md)** - IndieAuth Client Identification
- **[ADR-024](ADR-024-static-identity-page.md)** - Static Identity Page
- **[ADR-025](ADR-025-indieauth-pkce-authentication.md)** - IndieAuth PKCE Authentication
- **[ADR-026](ADR-026-indieauth-token-exchange-compliance.md)** - IndieAuth Token Exchange Compliance
- **[ADR-027](ADR-027-indieauth-authentication-endpoint-correction.md)** - IndieAuth Authentication Endpoint Correction
| ADR | Title | Status | Summary |
|---|---|---|---|
| [001](ADR-001-python-web-framework.md) | Python Web Framework Selection | Accepted | Flask as the web framework. |
| [002](ADR-002-flask-extensions.md) | Flask Extensions and Dependencies | Accepted | A minimal dependency set; avoid Flask extensions where the standard library will do. |
| [003](ADR-003-frontend-technology.md) | Front-end Technology Stack | Accepted | Server-rendered Jinja2 templates with minimal client-side code. |
| [006](ADR-006-python-virtual-environment-uv.md) | Python Virtual Environment Management with uv | Accepted | uv manages the virtual environment and dependencies. |
| [008](ADR-008-versioning-strategy.md) | Versioning Strategy | Accepted | Semantic Versioning, PEP 440 compatible. |
| [009](ADR-009-git-branching-strategy.md) | Git Branching Strategy | Accepted | Simplified trunk-based development with feature branches and release tags. |
| [012](ADR-012-flaky-test-removal.md) | Flaky Test Removal and Test Quality Standards | Accepted | Remove the multiprocessing migration tests and fix brittle assertions (done in v1.5.0). |
| [015](ADR-015-phase-5-implementation-approach.md) | Phase 5 Implementation Approach | Accepted | Version jump 0.5.1 → 0.6.0 for Phase 5 (historical). |
| [051](ADR-051-phase1-test-strategy.md) | Phase 1 Test Strategy and Implementation Review | Accepted | Review of the v1.1.1 Phase 1 test approach (historical). |
### Error Handling & Core Features (ADR-012 to ADR-015)
- **[ADR-012](ADR-012-http-error-handling-policy.md)** - HTTP Error Handling Policy
- **[ADR-013](ADR-013-expose-deleted-at-in-note-model.md)** - Expose Deleted-At in Note Model
- **[ADR-014](ADR-014-rss-feed-implementation.md)** - RSS Feed Implementation
- **[ADR-015](ADR-015-phase-5-implementation-approach.md)** - Phase 5 Implementation Approach
### Storage and database
### Micropub & API (ADR-028 to ADR-029)
- **[ADR-028](ADR-028-micropub-implementation.md)** - Micropub Implementation
- **[ADR-029](ADR-029-micropub-indieauth-integration.md)** - Micropub IndieAuth Integration
| ADR | Title | Status | Summary |
|---|---|---|---|
| [004](ADR-004-file-based-note-storage.md) | File-Based Note Storage Architecture | Accepted | Note content in Markdown files under `data/notes/YYYY/MM/`, metadata in SQLite. |
| [013](ADR-013-expose-deleted-at-in-note-model.md) | Expose deleted_at Field in Note Model | Accepted | `deleted_at` on the Note dataclass for soft deletes. |
| [020](ADR-020-automatic-database-migrations.md) | Automatic Database Migration System | Accepted | Numbered SQL migrations applied at startup and tracked in a table. |
| [031](ADR-031-database-migration-system-redesign.md) | Database Migration System Redesign | Accepted | Baseline schema first, then migrations, so existing databases upgrade (implemented via ADR-032/033). |
| [032](ADR-032-initial-schema-sql-implementation.md) | Initial Schema SQL Implementation for Migration System | Accepted | `INITIAL_SCHEMA_SQL` holds the baseline schema. |
| [033](ADR-033-database-migration-redesign.md) | Database Migration System Redesign | Accepted | Frozen baseline schema; every later change only in migration files (v1.1.0). |
| [037](ADR-037-migration-race-condition-fix.md) | Database Migration Race Condition Resolution | Accepted | SQLite `BEGIN IMMEDIATE` lock with retries so several gunicorn workers don't migrate at once. |
| [041](ADR-041-database-migration-conflict-resolution.md) | Database Migration Conflict Resolution | Accepted | Handling a production database that fell into a migration edge case. |
### Database & Migrations (ADR-020, ADR-031 to ADR-037)
- **[ADR-020](ADR-020-automatic-database-migrations.md)** - Automatic Database Migrations
- **[ADR-031](ADR-031-database-migration-system-redesign.md)** - Database Migration System Redesign
- **[ADR-032](ADR-032-initial-schema-sql-implementation.md)** - Initial Schema SQL Implementation
- **[ADR-033](ADR-033-database-migration-redesign.md)** - Database Migration Redesign
- **[ADR-037](ADR-037-migration-race-condition-fix.md)** - Migration Race Condition Fix
- **[ADR-041](ADR-041-database-migration-conflict-resolution.md)** - Database Migration Conflict Resolution
See also [database architecture](../architecture/database.md).
### Search & Advanced Features (ADR-034 to ADR-036, ADR-038 to ADR-040)
- **[ADR-034](ADR-034-full-text-search.md)** - Full-Text Search
- **[ADR-035](ADR-035-custom-slugs.md)** - Custom Slugs
- **[ADR-036](ADR-036-indieauth-token-verification-method.md)** - IndieAuth Token Verification Method
- **[ADR-038](ADR-038-syndication-formats.md)** - Syndication Formats (ATOM, JSON Feed)
- **[ADR-039](ADR-039-micropub-url-construction-fix.md)** - Micropub URL Construction Fix
- **[ADR-040](ADR-040-microformats2-compliance.md)** - Microformats2 Compliance
### Notes, slugs, media and search
### Architecture Refinements (ADR-042 to ADR-044)
- **[ADR-042](ADR-042-versioning-strategy-for-authorization-removal.md)** - Versioning Strategy for Authorization Removal
- **[ADR-043](ADR-043-CORRECTED-indieauth-endpoint-discovery.md)** - CORRECTED IndieAuth Endpoint Discovery
- **[ADR-044](ADR-044-endpoint-discovery-implementation.md)** - Endpoint Discovery Implementation Details
| ADR | Title | Status | Summary |
|---|---|---|---|
| [062](ADR-062-timestamp-based-slug-format.md) | Timestamp-Based Slug Format | Accepted* | Default slugs are `YYYYMMDDHHMMSS` timestamps (v1.5.0); replaces ADR-007. Articles take theirs from the title (ADR-077). |
| [035](ADR-035-custom-slugs.md) | Custom Slugs in Micropub | Accepted* | Honour `mp-slug`, sanitized, with reserved names protected (v1.1.0). |
| [034](ADR-034-full-text-search.md) | Full-Text Search with SQLite FTS5 | Accepted | FTS5 search over note content (v1.1.0). |
| [057](ADR-057-media-attachment-model.md) | Media Attachment Model | Accepted | Social-media-style image attachments: up to 4 per note, ordered, with captions. |
| [058](ADR-058-image-optimization-strategy.md) | Image Optimization Strategy | Accepted | Upload limits, automatic resizing and size variants. |
### Major Architectural Changes (ADR-050 to ADR-051)
- **[ADR-050](ADR-050-remove-custom-indieauth-server.md)** - Remove Custom IndieAuth Server
- **[ADR-051](ADR-051-phase1-test-strategy.md)** - Phase 1 Test Strategy
### Feeds
### v1.1.1 Quality & Production Readiness (ADR-052 to ADR-055)
- **[ADR-052](ADR-052-configuration-system-architecture.md)** - Configuration System Architecture
- **[ADR-053](ADR-053-performance-monitoring-strategy.md)** - Performance Monitoring Strategy
- **[ADR-054](ADR-054-structured-logging-architecture.md)** - Structured Logging Architecture
- **[ADR-055](ADR-055-error-handling-philosophy.md)** - Error Handling Philosophy
| ADR | Title | Status | Summary |
|---|---|---|---|
| [014](ADR-014-rss-feed-implementation.md) | RSS Feed Implementation Strategy | Accepted* | RSS 2.0 feed; its WebSub deferral is superseded by ADR-071, and item titles by ADR-076. |
| [038](ADR-038-syndication-formats.md) | Multiple Syndication Format Support | Accepted | RSS, Atom and JSON Feed side by side (v1.1.2). |
| [071](ADR-071-websub.md) | WebSub publishing with a built-in hub | Accepted | Built-in WebSub hub at `/websub` for StarPunk's own feeds (v2.10.0, amended v2.10.1). |
| [059](ADR-059-full-feed-media-standardization.md) | Full Feed Media Standardization (Option 3) | Proposed | Full Media RSS, podcasts, video and thumbnails in feeds; not scheduled. |
## ADRs by Topic
### IndieWeb: microformats, Webmention, Micropub
### Authentication & IndieAuth
ADR-005, ADR-010, ADR-011, ADR-016, ADR-017, ADR-018, ADR-019, ADR-021, ADR-022, ADR-023, ADR-024, ADR-025, ADR-026, ADR-027, ADR-036, ADR-043, ADR-044, ADR-050
| ADR | Title | Status | Summary |
|---|---|---|---|
| [040](ADR-040-microformats2-compliance.md) | Strict Microformats2 Compliance | Accepted* | Full h-entry, h-card and h-feed markup (v1.3.0, reviewed v2.8.0); `p-name`/`p-summary` per ADR-076. |
| [076](ADR-076-note-title-and-summary-columns.md) | Article titles and summaries stored apart from the text | Accepted | `notes.title`/`summary` columns; feeds title only articles; article previews and crossposts are title + summary (v2.16.0). |
| [077](ADR-077-article-urls-and-title-slugs.md) | Article URLs and slugs from titles | Accepted | Articles at `/article/<slug>` with slugs from their titles; the other prefix redirects (v2.17.0). |
| [061](ADR-061-author-discovery.md) | Author Profile Discovery from IndieAuth | Accepted | The author h-card is read from `ADMIN_ME` and cached. |
| [063](ADR-063-webmention-implementation.md) | Webmention Implementation | Accepted | Async receiving (202) and sending via a background worker, SSRF guard, rate limits, moderation (v1.6.0, revised up to v2.10.1). *Reconstructed.* |
| [028](ADR-028-micropub-implementation.md) | Micropub Implementation Strategy | Accepted* | `/micropub` endpoint for creating posts (v1.0.0-rc.1); tokens now per ADR-064. |
| [039](ADR-039-micropub-url-construction-fix.md) | Micropub URL Construction Fix | Accepted | Fix the double slash in returned post URLs. |
| [067](ADR-067-micropub-update-delete.md) | Micropub update, delete, drafts and replies | Accepted | Micropub update, delete, drafts and replies (v2.4.0). |
| [068](ADR-068-likes-and-reposts.md) | Likes and reposts | Accepted* | `like-of` and `repost-of` posts (v2.5.0); bookmarks were added later in v2.9.0. |
### Database & Migrations
ADR-004, ADR-020, ADR-031, ADR-032, ADR-033, ADR-037, ADR-041
See also [IndieWeb protocols](../architecture/indieweb-protocols.md).
### API & Micropub
ADR-028, ADR-029, ADR-039
### Crossposting (POSSE)
### Content & Features
ADR-007, ADR-013, ADR-014, ADR-034, ADR-035, ADR-038, ADR-040
| ADR | Title | Status | Summary |
|---|---|---|---|
| [065](ADR-065-direct-syndication-posse.md) | Direct Syndication (POSSE) to GoToSocial | Accepted | Crosspost to GoToSocial/Mastodon-API servers from a background worker (v2.1.0). |
| [066](ADR-066-bluesky-syndication.md) | Crossposting to Bluesky | Accepted | Bluesky as a second crosspost target (v2.2.0). |
| [069](ADR-069-threads-syndication.md) | Crossposting to Threads | Accepted | Threads as a third target, with automatic token refresh (v2.6.0). |
| [070](ADR-070-crosspost-links-and-cards.md) | Links and link cards in crossposts | Accepted | Keep note links in crossposts and build link cards for the linked page (v2.6.3). |
| [073](ADR-073-backfeed.md) | Backfeed of responses to crossposts | Accepted | Poll GoToSocial, Bluesky and Threads for responses to crossposts and store them as webmentions; admin backfill (v2.13.0, amended v2.13.2–v2.13.3). |
### Development & Operations
ADR-001, ADR-002, ADR-003, ADR-006, ADR-008, ADR-009, ADR-012, ADR-015, ADR-042, ADR-051, ADR-052, ADR-053, ADR-054, ADR-055
### Authentication (current)
## Superseded ADRs
| ADR | Title | Status | Summary |
|---|---|---|---|
| [064](ADR-064-oidc-admin-login-and-builtin-indieauth-server.md) | OIDC Admin Login and Built-in IndieAuth Server | Accepted | Admin login via OIDC; StarPunk issues and verifies Micropub tokens itself (v2.0.0). |
| [010](ADR-010-authentication-module-design.md) | Authentication Module Design | Accepted* | Single `auth.py` module and server-side sessions; the login flow is now ADR-064. |
| [011](ADR-011-development-authentication-mechanism.md) | Development Authentication Mechanism | Accepted | `DEV_MODE` login bypass for local development only. |
| [022](ADR-022-auth-route-prefix-fix.md) | Fix IndieAuth Callback Route Mismatch | Accepted | Auth routes live under `/auth` (v0.9.2). |
| [075](ADR-075-mcp-server.md) | MCP server for AI clients | Accepted | `/mcp` served by the official MCP SDK in a sidecar; StarPunk's IndieAuth server issues resource-bound tokens via client ID metadata documents (v2.15.0). |
These ADRs have been superseded by later decisions:
- **ADR-030** (old) - Superseded by ADR-043 (CORRECTED IndieAuth Endpoint Discovery)
### Operations, errors and monitoring
## How to Create a New ADR
| ADR | Title | Status | Summary |
|---|---|---|---|
| [012](ADR-012-http-error-handling-policy.md) | HTTP Error Handling Policy | Accepted | Correct HTTP status codes (e.g. 404 for missing notes) across routes. |
| [052](ADR-052-configuration-system-architecture.md) | Configuration System Architecture | Accepted | Environment variables with defaults, loaded in one module. |
| [054](ADR-054-structured-logging-architecture.md) | Structured Logging Architecture | Accepted | Rotating log files with correlation IDs (v1.1.1). |
| [055](ADR-055-error-handling-philosophy.md) | Error Handling Philosophy | Accepted | Layered handling: graceful degradation, clear user messages, detailed logs. |
| [072](ADR-072-remove-metrics-dashboard.md) | Remove the metrics dashboard | Accepted | Drop the in-memory metrics system and dashboard; keep `/health`, `/admin/health` and logs (v2.12.0). |
| [074](ADR-074-notifications.md) | Notifications through Apprise | Accepted | Notify the owner of new responses and crosspost/backfeed problems via the embedded Apprise library (v2.14.0). |
See also [deployment](../architecture/deployment.md).
## Superseded decisions
Kept for history. Most concern the IndieLogin.com and external-IndieAuth
designs of v0.x–v1.6.0, all replaced by ADR-064 in v2.0.0.
| ADR | Title | Status | Summary |
|---|---|---|---|
| [005](ADR-005-indielogin-authentication.md) | IndieLogin Authentication Integration | Superseded by 064 | Admin login through indielogin.com. |
| [007](ADR-007-slug-generation-algorithm.md) | Slug Generation Algorithm | Superseded by 062 | Slugs from the first words of the note. |
| [053](ADR-053-performance-monitoring-strategy.md) | Performance Monitoring Strategy | Superseded by 072 | Built-in, standard-library performance metrics (v1.1.1). |
| [060](ADR-060-production-hotfix-metrics-dashboard.md) | Route Adapter for the Metrics Dashboard | Superseded by 072 | Adapt monitoring data to the dashboard template in the route (v1.1.1-rc.2). *Reconstructed.* |
| [016](ADR-016-indieauth-client-discovery.md) | IndieAuth Client Discovery Mechanism | Superseded by 019 | h-app markup for client discovery. |
| [017](ADR-017-oauth-client-metadata-document.md) | OAuth Client ID Metadata Document Implementation | Superseded by 019 | OAuth client metadata endpoint. |
| [018](ADR-018-indieauth-detailed-logging.md) | IndieAuth Detailed Logging Strategy | Superseded by 064 | Redacted HTTP logging of the IndieAuth login flow. |
| [019](ADR-019-indieauth-correct-implementation.md) | IndieAuth Correct Implementation Based on IndieLogin.com API | Superseded by 025 | Draft of the IndieLogin.com PKCE flow. |
| [021](ADR-021-indieauth-provider-strategy.md) | IndieAuth Provider Strategy | Superseded by 064 | Keep IndieLogin.com for admin login. |
| [023](ADR-023-indieauth-client-identification.md) | IndieAuth Client Identification Strategy | Superseded by 064 | Visible h-app in the footer. |
| [024](ADR-024-static-identity-page.md) | Static HTML Identity Pages for IndieAuth | Superseded by 064 | Standalone identity page pointing at indieauth.com. |
| [025](ADR-025-indieauth-pkce-authentication.md) | IndieAuth Correct Implementation Based on IndieLogin.com API | Superseded by 064 | IndieLogin.com authentication with PKCE. |
| [026](ADR-026-indieauth-token-exchange-compliance.md) | IndieAuth Token Exchange Compliance | Superseded by 064 | Token exchange parameters for the IndieAuth login. |
| [027](ADR-027-indieauth-authentication-endpoint-correction.md) | IndieAuth Authentication Endpoint Correction | Superseded by 064 | Use the authentication (not token) endpoint for login. |
| [029](ADR-029-micropub-indieauth-integration.md) | Micropub IndieAuth Integration Strategy | Superseded by 050, 064 | Built-in token endpoint for Micropub (v1.0.0-rc.1). |
| [030](ADR-030-external-token-verification-architecture.md) | External Token Verification Architecture | Superseded by 043, 064 | Verify Micropub tokens at a configured external endpoint. |
| [036](ADR-036-indieauth-token-verification-method.md) | IndieAuth Token Verification Method Diagnosis | Superseded by 064 | How to call the external token endpoint. |
| [042](ADR-042-versioning-strategy-for-authorization-removal.md) | Versioning Strategy for Authorization Server Removal | Accepted (historical) | Remove the authorization server in 1.0.0-rc.4 rather than in a later release. |
| [043](ADR-043-CORRECTED-indieauth-endpoint-discovery.md) | IndieAuth Endpoint Discovery Architecture | Superseded by 064 | Discover the token endpoint from `ADMIN_ME` to verify tokens. |
| [044](ADR-044-endpoint-discovery-implementation.md) | IndieAuth Endpoint Discovery Implementation Details | Superseded by 064 | Implementation details of ADR-043. |
| [050](ADR-050-remove-custom-indieauth-server.md) | Remove Custom IndieAuth Server | Superseded by 064 | Remove the built-in authorization and token server (done in v1.0.0-rc.4). |
| [056](ADR-056-no-selfhosted-indieauth.md) | Use External IndieAuth Provider (Never Self-Host) | Superseded by 064 | Never run an IndieAuth server in StarPunk. |
ADR-042 is listed here because it only matters together with ADR-050; it
was carried out and is not itself superseded.
## Numbering
- **Gap ADR-045 to ADR-049**: never used; no document or code refers to
them.
- **Two ADR-012 files**: `ADR-012-http-error-handling-policy.md` and
`ADR-012-flaky-test-removal.md` (added 2025-12-16). Both are cited as
plain "ADR-012": `starpunk/routes/admin.py` and early CHANGELOG entries
mean the error policy, `tests/test_migration_race_condition.py` means
the flaky-test one. Not renumbered, since other documents link to them.
- **Old numbers in headings**: several files were renumbered without
updating their first heading: ADR-037 and ADR-038 say "ADR-022",
ADR-040 says "ADR-023", ADR-042 says "ADR-027", ADR-044 says "ADR-031".
Older CHANGELOG entries use those old numbers ("ADR-022: Migration Race
Condition Fix" is ADR-037; "ADR-031: IndieAuth Endpoint Discovery" is
ADR-044). The file name is authoritative.
- **ADR-060** was deleted in the v1.1.1-rc.2 review and reconstructed on
2026-11-03 because code still cites it. ADR-059 also mentions a planned
"Option 2 (ADR-060)" for Media RSS that was never written.
- **ADR-063** was lost with the uncommitted v1.6.0 docs and reconstructed
on 2026-11-03.
- **ADR-064 used twice**: v1.6.0 code (`starpunk/rate_limiter.py`,
`starpunk/routes/webmention.py`) cites a lost "ADR-064: Webmention Rate
Limiting"; the number was reused in v2.0.0 for OIDC login. The rate-limit
decision is recorded in ADR-063, Decision 10.
## How to create a new ADR
1. Use the next number after the highest existing one (currently ADR-077).
2. Name it `ADR-NNN-brief-descriptive-title.md`.
3. Follow the format of recent ADRs (e.g. ADR-070, ADR-071):
1. **Find the next sequential number**: Check the highest existing ADR number
2. **Use the naming format**: `ADR-NNN-brief-descriptive-title.md`
3. **Follow the template**:
```markdown
# ADR-NNN: Title
## Status
Proposed | Accepted | Deprecated | Superseded
Accepted (vX.Y.Z, YYYY-MM-DD)
## Context
Why are we making this decision?
## Decision
What have we decided to do?
## Consequences
What are the positive and negative consequences?
## Alternatives Considered
What other options did we evaluate?
## Alternatives considered
```
4. **Update this index** with the new ADR
## Related Documentation
- **[../architecture/](../architecture/)** - Architectural overviews and system design
- **[../design/](../design/)** - Detailed design documents
- **[../standards/](../standards/)** - Coding standards and conventions
4. When a decision replaces an earlier one, change the earlier ADR's Status
to "Superseded by ADR-NNN" with a one-line note; don't rewrite its body.
5. Add the ADR to this index.
## Related documentation
- [../architecture/](../architecture/) — current system design
- [../design/](../design/) — per-version implementation designs
- [../standards/](../standards/) — coding and documentation standards
- [../../CHANGELOG.md](../../CHANGELOG.md) — what changed in each release
---
**Last Updated**: 2025-11-25
**Maintained By**: Documentation Manager Agent
**Total ADRs**: 55
**Last Updated**: 2026-11-11
**Total ADR files**: 72
-41
View File
@@ -1,41 +0,0 @@
# Deployment Documentation Index
This directory contains deployment guides, infrastructure setup instructions, and operations documentation for StarPunk CMS.
## Deployment Guides
- **[container-deployment.md](container-deployment.md)** - Container-based deployment guide (Docker, Podman)
## Deployment Options
### Container Deployment (Recommended)
Container deployment provides:
- Consistent environment across platforms
- Easy updates and rollbacks
- Resource isolation
- Simplified dependency management
See: [container-deployment.md](container-deployment.md)
### Manual Deployment
For manual deployment without containers:
- Follow [../standards/development-setup.md](../standards/development-setup.md)
- Configure systemd service
- Set up reverse proxy (nginx/Caddy)
- Configure SSL/TLS certificates
### Cloud Deployment
StarPunk can be deployed to:
- Any container platform (Kubernetes, Docker Swarm)
- VPS providers (DigitalOcean, Linode, Vultr)
- PaaS platforms supporting containers
## Related Documentation
- **[../standards/development-setup.md](../standards/development-setup.md)** - Development environment setup
- **[../architecture/](../architecture/)** - System architecture
- **[README.md](../../README.md)** - Quick start guide
---
**Last Updated**: 2025-11-25
**Maintained By**: Documentation Manager Agent
+450 -96
View File
@@ -1,128 +1,482 @@
# Design Documentation Index
This directory contains detailed design documents, feature specifications, and phase implementation plans for StarPunk CMS.
Implementation designs, developer questions and architect answers,
implementation reports (`YYYY-MM-DD-description.md`) and reviews, one
folder per release (see `docs/DOCUMENTATION.md`). Current system design
is in `../architecture/`; decisions are in `../decisions/` (ADRs).
## Project Structure
- **[project-structure.md](project-structure.md)** - Overall project structure and organization
- **[initial-files.md](initial-files.md)** - Initial file structure for the project
## Notes
## Phase Implementation Plans
- v1.6.0 (webmentions) was deployed but never committed; its code was
recovered from the production image on 2026-10-31 (commit `d3fa5b7`). Its
documents and tests were not in the image. The five v1.6.0 documents the
code cites were reconstructed from the implementation.
- Releases from v2.0.1 on that had no folder have a short implementation
report reconstructed from the changelog and git history.
- Documents marked *(reconstructed)* were written on 2026-11-03 after the fact.
- Every release from v2.0.0 on has a folder. Of the v1 releases, v1.0.1
and v1.1.0 have none: their documents are in `v1.1.1/` (and the v1.0.1
hotfix plan in `v1.0.0/`).
- Description = the document's title.
### Phase 1: Foundation
- **[phase-1.1-core-utilities.md](phase-1.1-core-utilities.md)** - Core utility functions and helpers
- **[phase-1.1-quick-reference.md](phase-1.1-quick-reference.md)** - Quick reference for Phase 1.1
- **[phase-1.2-data-models.md](phase-1.2-data-models.md)** - Data models and database schema
- **[phase-1.2-quick-reference.md](phase-1.2-quick-reference.md)** - Quick reference for Phase 1.2
## Contents
### Phase 2: Core Features
- **[phase-2.1-notes-management.md](phase-2.1-notes-management.md)** - Notes CRUD functionality
- **[phase-2.1-quick-reference.md](phase-2.1-quick-reference.md)** - Quick reference for Phase 2.1
[v0.x](#v0x) · [v1.0.0](#v100) · [v1.1.1](#v111) · [v1.1.2](#v112) · [v1.2.0](#v120) · [v1.3.0](#v130) · [v1.3.1](#v131) · [v1.4.0](#v140) · [v1.4.1](#v141) · [v1.4.2](#v142) · [v1.5.0](#v150) · [v1.6.0](#v160) · [v2.0.0](#v200) · [v2.0.1](#v201) · [v2.0.2](#v202) · [v2.1.0](#v210) · [v2.2.0](#v220) · [v2.3.0](#v230) · [v2.3.1](#v231) · [v2.3.2](#v232) · [v2.3.3](#v233) · [v2.4.0](#v240) · [v2.5.0](#v250) · [v2.6.0](#v260) · [v2.6.1](#v261) · [v2.6.2](#v262) · [v2.6.3](#v263) · [v2.6.4](#v264) · [v2.6.5](#v265) · [v2.6.6](#v266) · [v2.6.7](#v267) · [v2.6.8](#v268) · [v2.6.9](#v269) · [v2.7.0](#v270) · [v2.7.1](#v271) · [v2.7.2](#v272) · [v2.8.0](#v280) · [v2.8.1](#v281) · [v2.9.0](#v290) · [v2.10.0](#v2100) · [v2.10.1](#v2101) · [v2.10.2](#v2102) · [v2.11.0](#v2110) · [v2.12.0](#v2120) · [v2.12.1](#v2121) · [v2.13.0](#v2130) · [v2.13.1](#v2131) · [v2.13.2](#v2132) · [v2.13.3](#v2133) · [v2.13.4](#v2134) · [v2.14.0](#v2140) · [v2.14.1](#v2141) · [v2.14.2](#v2142) · [hotfix](#hotfix)
### Phase 3: Authentication
- **[phase-3-authentication.md](phase-3-authentication.md)** - Authentication system design
- **[phase-3-authentication-implementation.md](phase-3-authentication-implementation.md)** - Implementation details
- **[indieauth-pkce-authentication.md](indieauth-pkce-authentication.md)** - IndieAuth PKCE authentication design
## v0.x
### Phase 4: Web Interface
- **[phase-4-web-interface.md](phase-4-web-interface.md)** - Web interface design
- **[phase-4-quick-reference.md](phase-4-quick-reference.md)** - Quick reference for Phase 4
- **[phase-4-error-handling-fix.md](phase-4-error-handling-fix.md)** - Error handling improvements
Not a release folder: documents from before v1.0.0 kept for history.
### Phase 5: RSS & Deployment
- **[phase-5-rss-and-container.md](phase-5-rss-and-container.md)** - RSS feed and container deployment
- **[phase-5-executive-summary.md](phase-5-executive-summary.md)** - Executive summary of Phase 5
- **[phase-5-quick-reference.md](phase-5-quick-reference.md)** - Quick reference for Phase 5
- [technology-stack-legacy.md](v0.x/technology-stack-legacy.md) — StarPunk Technology Stack - Quick Reference
## Feature-Specific Design
## v1.0.0
### Micropub API
- **[micropub-endpoint-design.md](micropub-endpoint-design.md)** - Micropub endpoint detailed design
V1 foundation: phases 1–5 (utilities, models, notes, auth, web UI, RSS, container), Micropub, migrations, IndieAuth fixes and the release candidates.
### Authentication Fixes
- **[auth-redirect-loop-diagnosis.md](auth-redirect-loop-diagnosis.md)** - Diagnosis of redirect loop issues
- **[auth-redirect-loop-diagram.md](auth-redirect-loop-diagram.md)** - Visual diagrams of the problem
- **[auth-redirect-loop-executive-summary.md](auth-redirect-loop-executive-summary.md)** - Executive summary
- **[auth-redirect-loop-fix-implementation.md](auth-redirect-loop-fix-implementation.md)** - Implementation guide
- [2025-11-18-auth-redirect-loop-fix.md](v1.0.0/2025-11-18-auth-redirect-loop-fix.md) — Auth Redirect Loop Fix - Implementation Report
- [2025-11-18-quickfix-auth-loop.md](v1.0.0/2025-11-18-quickfix-auth-loop.md) — QUICK FIX: Auth Redirect Loop
- [2025-11-19-container-implementation-summary.md](v1.0.0/2025-11-19-container-implementation-summary.md) — Phase 5 Containerization - Implementation Complete
- [2025-11-19-migration-implementation-quick-reference.md](v1.0.0/2025-11-19-migration-implementation-quick-reference.md) — Migration System - Quick Reference Card
- [2025-11-19-migration-system-implementation-guidance.md](v1.0.0/2025-11-19-migration-system-implementation-guidance.md) — Migration System Implementation Guidance
- [2025-11-19-migration-system-implementation-report.md](v1.0.0/2025-11-19-migration-system-implementation-report.md) — Migration System Implementation Report
- [2025-11-19-todo-test-updates.md](v1.0.0/2025-11-19-todo-test-updates.md) — Test Updates Required for ADR-019 Implementation
- [2025-11-22-auth-route-prefix-fix.md](v1.0.0/2025-11-22-auth-route-prefix-fix.md) — Auth Route Prefix Fix Implementation Report
- [2025-11-22-authorization-endpoint-fix.md](v1.0.0/2025-11-22-authorization-endpoint-fix.md) — IndieAuth Authentication Endpoint Correction
- [2025-11-22-grant-type-fix.md](v1.0.0/2025-11-22-grant-type-fix.md) — IndieAuth Token Exchange grant_type Fix
- [2025-11-24-endpoint-discovery-analysis.md](v1.0.0/2025-11-24-endpoint-discovery-analysis.md) — IndieAuth Endpoint Discovery Implementation Analysis
- [2025-11-24-indieauth-removal-complete.md](v1.0.0/2025-11-24-indieauth-removal-complete.md) — IndieAuth Server Removal - Complete Implementation Report
- [2025-11-24-migration-detection-hotfix-rc3.md](v1.0.0/2025-11-24-migration-detection-hotfix-rc3.md) — Migration Detection Hotfix - v1.0.0-rc.3
- [2025-11-24-migration-fix-v1.0.0-rc.2.md](v1.0.0/2025-11-24-migration-fix-v1.0.0-rc.2.md) — Implementation Report: Migration Fix for v1.0.0-rc.2
- [2025-11-24-phase1-indieauth-server-removal.md](v1.0.0/2025-11-24-phase1-indieauth-server-removal.md) — Phase 1: IndieAuth Authorization Server Removal - Implementation Report
- [2025-11-24-v1.0.0-rc.5-implementation.md](v1.0.0/2025-11-24-v1.0.0-rc.5-implementation.md) — v1.0.0-rc.5 Implementation Report
- [ADR-025-implementation-report.md](v1.0.0/ADR-025-implementation-report.md) — ADR-025 Implementation Report
- [ADR-025-implementation-summary.md](v1.0.0/ADR-025-implementation-summary.md) — ADR-025 Implementation Summary
- [ADR-025-versioning-guidance.md](v1.0.0/ADR-025-versioning-guidance.md) — ADR-019 Implementation: Versioning Guidance
- [ARCHITECT-FINAL-ANALYSIS.md](v1.0.0/ARCHITECT-FINAL-ANALYSIS.md) — Architect Final Analysis - Delete Route 404 Fix
- [container-deployment.md](v1.0.0/container-deployment.md) — StarPunk Container Deployment Guide
- [custom-slug-bug-diagnosis.md](v1.0.0/custom-slug-bug-diagnosis.md) — Custom Slug Bug Diagnosis Report
- [custom-slug-bug-implementation.md](v1.0.0/custom-slug-bug-implementation.md) — Custom Slug Bug Fix - Implementation Report
- [database-migration-architecture.md](v1.0.0/database-migration-architecture.md) — Database Migration Architecture
- [database-migration-conflict-diagnosis.md](v1.0.0/database-migration-conflict-diagnosis.md) — Database Migration Conflict Diagnosis Report
- [delete-nonexistent-note-error-analysis.md](v1.0.0/delete-nonexistent-note-error-analysis.md) — Delete Nonexistent Note Error Analysis
- [delete-route-404-fix-implementation.md](v1.0.0/delete-route-404-fix-implementation.md) — Delete Route 404 Fix - Implementation Report
- [delete-route-fix-summary.md](v1.0.0/delete-route-fix-summary.md) — Delete Route Fix - Developer Summary
- [delete-route-implementation-spec.md](v1.0.0/delete-route-implementation-spec.md) — Delete Route Implementation Specification
- [dependencies-diagram.md](v1.0.0/dependencies-diagram.md) — StarPunk V1 Dependencies Diagram
- [endpoint-discovery-answers.md](v1.0.0/endpoint-discovery-answers.md) — IndieAuth Endpoint Discovery: Definitive Implementation Answers
- [error-handling-rest-vs-web-patterns.md](v1.0.0/error-handling-rest-vs-web-patterns.md) — Architectural Review: Error Handling in Web Routes
- [feature-scope.md](v1.0.0/feature-scope.md) — StarPunk V1 Feature Scope
- [fix-hardcoded-endpoints.md](v1.0.0/fix-hardcoded-endpoints.md) — Migration Guide: Fixing Hardcoded IndieAuth Endpoints
- [identity-domain-validation-2025-11-19.md](v1.0.0/identity-domain-validation-2025-11-19.md) — Identity Domain Validation Report
- [identity-page-customization-guide.md](v1.0.0/identity-page-customization-guide.md) — IndieAuth Identity Page Customization Guide
- [identity-page.html](v1.0.0/identity-page.html) — Example IndieAuth identity page (HTML template)
- [implementation-guide-expose-deleted-at.md](v1.0.0/implementation-guide-expose-deleted-at.md) — Implementation Guide: Expose deleted_at in Note Model
- [implementation-plan.md](v1.0.0/implementation-plan.md) — StarPunk V1 Implementation Plan
- [indieauth-assessment.md](v1.0.0/indieauth-assessment.md) — IndieAuth Architecture Assessment
- [indieauth-client-diagnosis.md](v1.0.0/indieauth-client-diagnosis.md) — IndieAuth Client Registration Issue - Diagnosis Report
- [indieauth-client-discovery-analysis.md](v1.0.0/indieauth-client-discovery-analysis.md) — IndieAuth Client Discovery Error - Architectural Analysis
- [indieauth-client-discovery-fix-implementation.md](v1.0.0/indieauth-client-discovery-fix-implementation.md) — IndieAuth Client Discovery Fix - Implementation Report
- [indieauth-client-discovery-root-cause-analysis.md](v1.0.0/indieauth-client-discovery-root-cause-analysis.md) — IndieAuth Client Discovery Root Cause Analysis
- [indieauth-detailed-logging-implementation.md](v1.0.0/indieauth-detailed-logging-implementation.md) — IndieAuth Detailed Logging Implementation Report
- [indieauth-endpoint-discovery-security.md](v1.0.0/indieauth-endpoint-discovery-security.md) — IndieAuth Endpoint Discovery Security Analysis
- [indieauth-endpoint-discovery.md](v1.0.0/indieauth-endpoint-discovery.md) — IndieAuth Endpoint Discovery Architecture
- [indieauth-fix-summary.md](v1.0.0/indieauth-fix-summary.md) — IndieAuth Authentication Fix - Quick Summary
- [indieauth-identity-page.md](v1.0.0/indieauth-identity-page.md) — IndieAuth Identity Page Architecture
- [indieauth-questions-answered.md](v1.0.0/indieauth-questions-answered.md) — IndieAuth Implementation Questions - Answered
- [indieauth-removal-analysis.md](v1.0.0/indieauth-removal-analysis.md) — IndieAuth Removal Implementation Analysis
- [indieauth-removal-architectural-review.md](v1.0.0/indieauth-removal-architectural-review.md) — Architectural Review: IndieAuth Authorization Server Removal
- [indieauth-removal-implementation-guide.md](v1.0.0/indieauth-removal-implementation-guide.md) — IndieAuth Provider Removal - Implementation Guide
- [indieauth-removal-phases.md](v1.0.0/indieauth-removal-phases.md) — IndieAuth Removal: Phased Implementation Guide
- [indieauth-removal-plan.md](v1.0.0/indieauth-removal-plan.md) — IndieAuth Server Removal Plan
- [indieauth-removal-questions.md](v1.0.0/indieauth-removal-questions.md) — IndieAuth Removal - Questions for Architect
- [indieauth-spec-url-standardization-2025-11-24.md](v1.0.0/indieauth-spec-url-standardization-2025-11-24.md) — IndieAuth Specification URL Standardization Report
- [indieauth-token-verification-diagnosis.md](v1.0.0/indieauth-token-verification-diagnosis.md) — IndieAuth Token Verification Diagnosis
- [initial-files.md](v1.0.0/initial-files.md) — Initial Project Files Design
- [initial-schema-implementation-guide.md](v1.0.0/initial-schema-implementation-guide.md) — Initial Schema SQL Implementation Guide
- [initial-schema-quick-reference.md](v1.0.0/initial-schema-quick-reference.md) — INITIAL_SCHEMA_SQL Quick Reference
- [micropub-401-diagnosis.md](v1.0.0/micropub-401-diagnosis.md) — Micropub 401 Unauthorized Error - Architectural Diagnosis
- [micropub-endpoint-design.md](v1.0.0/micropub-endpoint-design.md) — Micropub Endpoint Design for StarPunk V1
- [micropub-phase1-architecture-review.md](v1.0.0/micropub-phase1-architecture-review.md) — Micropub V1 Implementation - Phase 1 Architecture Review
- [micropub-phase3-architecture-review.md](v1.0.0/micropub-phase3-architecture-review.md) — Micropub Phase 3 Implementation Architecture Review
- [micropub-v1-implementation-progress.md](v1.0.0/micropub-v1-implementation-progress.md) — Micropub V1 Implementation Progress Report
- [migration-failure-diagnosis-v1.0.0-rc.1.md](v1.0.0/migration-failure-diagnosis-v1.0.0-rc.1.md) — Migration Failure Diagnosis - v1.0.0-rc.1
- [migration-fix-quick-reference.md](v1.0.0/migration-fix-quick-reference.md) — Migration Race Condition Fix - Quick Implementation Reference
- [migration-race-condition-answers.md](v1.0.0/migration-race-condition-answers.md) — Migration Race Condition Fix - Architectural Answers
- [migration-race-condition-fix-implementation.md](v1.0.0/migration-race-condition-fix-implementation.md) — Migration Race Condition Fix - Implementation Guide
- [oauth-metadata-implementation-2025-11-19.md](v1.0.0/oauth-metadata-implementation-2025-11-19.md) — OAuth Client ID Metadata Document Implementation Report
- [phase-1.1-core-utilities.md](v1.0.0/phase-1.1-core-utilities.md) — Phase 1.1: Core Utilities Design
- [phase-1.1-quick-reference.md](v1.0.0/phase-1.1-quick-reference.md) — Phase 1.1 Quick Reference: Core Utilities
- [phase-1.2-data-models.md](v1.0.0/phase-1.2-data-models.md) — Phase 1.2: Data Models Design
- [phase-1.2-quick-reference.md](v1.0.0/phase-1.2-quick-reference.md) — Phase 1.2 Quick Reference: Data Models
- [phase-2-architectural-review.md](v1.0.0/phase-2-architectural-review.md) — Architectural Review: Phase 2 Implementation
- [phase-2-implementation-report.md](v1.0.0/phase-2-implementation-report.md) — Phase 2 Implementation Report: Authorization and Token Endpoints
- [phase-2.1-implementation-20251118.md](v1.0.0/phase-2.1-implementation-20251118.md) — Phase 2.1 Implementation Report: Notes Management
- [phase-2.1-notes-management.md](v1.0.0/phase-2.1-notes-management.md) — Phase 2.1: Notes Management Design
- [phase-2.1-quick-reference.md](v1.0.0/phase-2.1-quick-reference.md) — Phase 2.1: Notes Management - Quick Reference
- [phase-3-authentication-20251118.md](v1.0.0/phase-3-authentication-20251118.md) — Phase 3: Authentication Implementation Report
- [phase-3-authentication-architectural-review.md](v1.0.0/phase-3-authentication-architectural-review.md) — Phase 3: Authentication Implementation - Architectural Review
- [phase-3-authentication-implementation.md](v1.0.0/phase-3-authentication-implementation.md) — Phase 3: Authentication Implementation Design
- [phase-3-authentication.md](v1.0.0/phase-3-authentication.md) — Phase 3: Authentication Design
- [phase-4-architectural-assessment-20251118.md](v1.0.0/phase-4-architectural-assessment-20251118.md) — Phase 4: Architectural Assessment and Design Report
- [phase-4-error-handling-fix.md](v1.0.0/phase-4-error-handling-fix.md) — Phase 4: Error Handling Fix - Implementation Guide
- [phase-4-quick-reference.md](v1.0.0/phase-4-quick-reference.md) — Phase 4: Quick Reference
- [phase-4-test-fixes.md](v1.0.0/phase-4-test-fixes.md) — Phase 4 Test Fixes Report
- [phase-4-web-interface.md](v1.0.0/phase-4-web-interface.md) — Phase 4: Web Interface - Implementation Design
- [phase-5-approval-summary.md](v1.0.0/phase-5-approval-summary.md) — Phase 5 Containerization - Approval Summary
- [phase-5-container-architectural-review.md](v1.0.0/phase-5-container-architectural-review.md) — Phase 5 Containerization - Architectural Review
- [phase-5-container-implementation-report.md](v1.0.0/phase-5-container-implementation-report.md) — Phase 5 Container Implementation Report
- [phase-5-executive-summary.md](v1.0.0/phase-5-executive-summary.md) — Phase 5 Executive Summary
- [phase-5-pre-implementation-review.md](v1.0.0/phase-5-pre-implementation-review.md) — Phase 5 Pre-Implementation Review
- [phase-5-quick-reference.md](v1.0.0/phase-5-quick-reference.md) — Phase 5 Quick Reference Guide
- [phase-5-rss-and-container.md](v1.0.0/phase-5-rss-and-container.md) — Phase 5: RSS Feed & Production Container - Implementation Design
- [phase-5-rss-implementation-20251119.md](v1.0.0/phase-5-rss-implementation-20251119.md) — Phase 5: RSS Feed Implementation Report
- [phase-5-validation-report.md](v1.0.0/phase-5-validation-report.md) — Phase 5 RSS Feed Implementation - Architectural Validation Report
- [phase1-completion-guide.md](v1.0.0/phase1-completion-guide.md) — Phase 1 Completion Guide: Test Cleanup and Commit
- [project-structure.md](v1.0.0/project-structure.md) — Project Structure Design
- [quick-reference.md](v1.0.0/quick-reference.md) — StarPunk V1 Quick Reference
- [review-v1.0.0-rc.5.md](v1.0.0/review-v1.0.0-rc.5.md) — Architectural Review: v1.0.0-rc.5 Implementation
- [setup-complete-2025-11-18.md](v1.0.0/setup-complete-2025-11-18.md) — StarPunk Project Setup - Complete
- [test-failure-analysis-deleted-at-attribute.md](v1.0.0/test-failure-analysis-deleted-at-attribute.md) — Test Failure Analysis: Missing `deleted_at` Attribute on Note Model
- [troubleshooting.md](v1.0.0/troubleshooting.md) — StarPunk Troubleshooting Guide
- [v0.9.1-implementation-report.md](v1.0.0/v0.9.1-implementation-report.md) — StarPunk v0.9.1 Implementation Report
- [v1-planning-overview.md](v1.0.0/v1-planning-overview.md) — StarPunk V1 Project Plan
- [v1.0.0-rc.1-hotfix-instructions.md](v1.0.0/v1.0.0-rc.1-hotfix-instructions.md) — v1.0.0-rc.1 Production Hotfix Instructions
- [v1.0.0-rc.5-migration-race-condition-implementation.md](v1.0.0/v1.0.0-rc.5-migration-race-condition-implementation.md) — v1.0.0-rc.5 Migration Race Condition Fix - Implementation Report
- [v1.0.0-release-validation.md](v1.0.0/v1.0.0-release-validation.md) — StarPunk v1.0.0 Release Validation Report
- [v1.0.1-hotfix-plan.md](v1.0.0/v1.0.1-hotfix-plan.md) — StarPunk v1.0.1 Hotfix Release Plan
### Database Schema
- **[initial-schema-implementation-guide.md](initial-schema-implementation-guide.md)** - Schema implementation guide
- **[initial-schema-quick-reference.md](initial-schema-quick-reference.md)** - Quick reference
## v1.1.1
### Security
- **[token-security-migration.md](token-security-migration.md)** - Token security improvements
v1.1.0 "SearchLight" (search) and v1.1.1 "Polish" designs and reports; v1.0.1 hotfix.
## Version-Specific Design
- [2025-11-25-hotfix-v1.1.1-rc.2-implementation.md](v1.1.1/2025-11-25-hotfix-v1.1.1-rc.2-implementation.md) — Implementation Report: Hotfix v1.1.1-rc.2 - Admin Dashboard Route Conflict
- [2025-11-25-v1.0.1-micropub-url-fix.md](v1.1.1/2025-11-25-v1.0.1-micropub-url-fix.md) — v1.0.1 Hotfix Implementation Report
- [auth-redirect-loop-diagnosis.md](v1.1.1/auth-redirect-loop-diagnosis.md) — Authentication Redirect Loop Diagnosis - Phase 4
- [auth-redirect-loop-diagram.md](v1.1.1/auth-redirect-loop-diagram.md) — Auth Redirect Loop - Visual Diagram
- [auth-redirect-loop-executive-summary.md](v1.1.1/auth-redirect-loop-executive-summary.md) — Auth Redirect Loop - Executive Summary
- [auth-redirect-loop-fix-implementation.md](v1.1.1/auth-redirect-loop-fix-implementation.md) — Implementation Guide: Auth Redirect Loop Fix
- [hotfix-v1.1.1-rc2-consolidated.md](v1.1.1/hotfix-v1.1.1-rc2-consolidated.md) — Hotfix Design: v1.1.1-rc.2 - Metrics Dashboard Template Data Mismatch
- [hotfix-v1.1.1-rc2-review.md](v1.1.1/hotfix-v1.1.1-rc2-review.md) — Architectural Review: Hotfix v1.1.1-rc.2
- [hotfix-v1.1.1-rc2-route-conflict.md](v1.1.1/hotfix-v1.1.1-rc2-route-conflict.md) — Hotfix Design: v1.1.1-rc.2 Route Conflict Resolution
- [hotfix-validation-script.md](v1.1.1/hotfix-validation-script.md) — Hotfix Validation Script for v1.1.1-rc.2
- [indieauth-pkce-authentication.md](v1.1.1/indieauth-pkce-authentication.md) — IndieAuth PKCE Authentication - Technical Design
- [potential-features.md](v1.1.1/potential-features.md) — StarPunk V1.1: Potential Features
- [priority-work.md](v1.1.1/priority-work.md) — StarPunk v1.1.0: Priority Work Items
- [syndication-features.md](v1.1.1/syndication-features.md) — Syndication Features Specification
- [token-security-migration.md](v1.1.1/token-security-migration.md) — Token Security Migration Strategy
- [upgrade-to-v1.1.1.md](v1.1.1/upgrade-to-v1.1.1.md) — Upgrade Guide: StarPunk v1.1.1 "Polish"
- [v1.1.0-feature-architecture.md](v1.1.1/v1.1.0-feature-architecture.md) — StarPunk v1.1.0 Feature Architecture
- [v1.1.0-implementation-decisions.md](v1.1.1/v1.1.0-implementation-decisions.md) — V1.1.0 Implementation Decisions - Architectural Guidance
- [v1.1.0-implementation-plan.md](v1.1.1/v1.1.0-implementation-plan.md) — StarPunk v1.1.0 Implementation Plan
- [v1.1.0-implementation-report.md](v1.1.1/v1.1.0-implementation-report.md) — StarPunk v1.1.0 Implementation Report
- [v1.1.0-release-status.md](v1.1.1/v1.1.0-release-status.md) — StarPunk v1.1.0 "SearchLight" Release Status
- [v1.1.0-search-ui-validation.md](v1.1.1/v1.1.0-search-ui-validation.md) — StarPunk v1.1.0 Search UI Implementation Review
- [v1.1.0-validation-report.md](v1.1.1/v1.1.0-validation-report.md) — StarPunk v1.1.0 Implementation Validation & Search UI Design
- [v1.1.1-architecture-overview.md](v1.1.1/v1.1.1-architecture-overview.md) — v1.1.1 "Polish" Architecture Overview
- [v1.1.1-final-release-review.md](v1.1.1/v1.1.1-final-release-review.md) — StarPunk v1.1.1 "Polish" - Final Architectural Release Review
- [v1.1.1-instrumentation-assessment.md](v1.1.1/v1.1.1-instrumentation-assessment.md) — v1.1.1 Performance Monitoring Instrumentation Assessment
- [v1.1.1-phase1-architectural-review.md](v1.1.1/v1.1.1-phase1-architectural-review.md) — StarPunk v1.1.1 Phase 1 - Architectural Review Report
- [v1.1.1-phase1-implementation.md](v1.1.1/v1.1.1-phase1-implementation.md) — StarPunk v1.1.1 Phase 1 Implementation Report
- [v1.1.1-phase2-architectural-review.md](v1.1.1/v1.1.1-phase2-architectural-review.md) — StarPunk v1.1.1 "Polish" - Phase 2 Architectural Review
- [v1.1.1-phase2-implementation.md](v1.1.1/v1.1.1-phase2-implementation.md) — StarPunk v1.1.1 "Polish" - Phase 2 Implementation Report
- [v1.1.1-phase3-implementation.md](v1.1.1/v1.1.1-phase3-implementation.md) — StarPunk v1.1.1 "Polish" - Phase 3 Implementation Report
### v1.1.1
- **[v1.1.1/](v1.1.1/)** - v1.1.1 specific design documents
## v1.1.2
## Quick Reference Documents
v1.1.2 "Syndicate": metrics, ATOM/JSON feeds, feed statistics, OPML.
Quick reference documents provide condensed, actionable information for developers:
- **phase-1.1-quick-reference.md** - Core utilities quick ref
- **phase-1.2-quick-reference.md** - Data models quick ref
- **phase-2.1-quick-reference.md** - Notes management quick ref
- **phase-4-quick-reference.md** - Web interface quick ref
- **phase-5-quick-reference.md** - RSS and deployment quick ref
- **initial-schema-quick-reference.md** - Database schema quick ref
- [2025-11-26-phase2-architect-review.md](v1.1.2/2025-11-26-phase2-architect-review.md) — Architectural Review: StarPunk v1.1.2 Phase 2 "Syndicate" - Feed Formats
- [2025-11-26-v1.1.2-phase1-review.md](v1.1.2/2025-11-26-v1.1.2-phase1-review.md) — StarPunk v1.1.2 Phase 1 Implementation Review
- [2025-11-26-v1.1.2-phase2-complete.md](v1.1.2/2025-11-26-v1.1.2-phase2-complete.md) — StarPunk v1.1.2 Phase 2 Feed Formats - Implementation Report (COMPLETE)
- [2025-11-26-v1.1.2-phase2-feed-formats-partial.md](v1.1.2/2025-11-26-v1.1.2-phase2-feed-formats-partial.md) — StarPunk v1.1.2 Phase 2 Feed Formats - Implementation Report (Partial)
- [2025-11-27-phase3-architect-review.md](v1.1.2/2025-11-27-phase3-architect-review.md) — StarPunk v1.1.2 Phase 3 - Architectural Review
- [2025-11-27-v1.1.2-phase3-complete.md](v1.1.2/2025-11-27-v1.1.2-phase3-complete.md) — v1.1.2 Phase 3 Implementation Report - Feed Statistics & OPML
- [2025-11-28-v1.1.2-rc.1-architect-review.md](v1.1.2/2025-11-28-v1.1.2-rc.1-architect-review.md) — Architect Review: v1.1.2-rc.1 Production Issues
- [2025-11-28-v1.1.2-rc.1-production-issues.md](v1.1.2/2025-11-28-v1.1.2-rc.1-production-issues.md) — v1.1.2-rc.1 Production Issues Investigation Report
- [2025-11-28-v1.1.2-rc.2-fixes.md](v1.1.2/2025-11-28-v1.1.2-rc.2-fixes.md) — v1.1.2-rc.2 Production Bug Fixes - Implementation Report
- [feed-media-handling-options.md](v1.1.2/feed-media-handling-options.md) — Feed Media Handling: Architecture Options Analysis
- [feed-media-option2-design.md](v1.1.2/feed-media-option2-design.md) — Feed Media Enhancement Design: Option 2 (RSS + Media RSS Extension)
- [upgrade-to-v1.1.2.md](v1.1.2/upgrade-to-v1.1.2.md) — Upgrade Guide: StarPunk v1.1.2 "Syndicate"
- [v1.1.2-caption-alttext-update.md](v1.1.2/v1.1.2-caption-alttext-update.md) — Caption Display Update - Alt Text Only (v1.1.2)
- [v1.1.2-options.md](v1.1.2/v1.1.2-options.md) — StarPunk v1.1.2 Release Plan Options
- [v1.1.2-phase1-metrics-implementation.md](v1.1.2/v1.1.2-phase1-metrics-implementation.md) — StarPunk v1.1.2 Phase 1: Metrics Instrumentation - Implementation Report
- [v1.1.2-syndicate-architecture.md](v1.1.2/v1.1.2-syndicate-architecture.md) — StarPunk v1.1.2 "Syndicate" - Architecture Overview
## How to Use This Documentation
## v1.2.0
### For Developers Implementing Features
1. Start with the relevant **phase** document (e.g., phase-2.1-notes-management.md)
2. Consult the **quick reference** for that phase
3. Check **feature-specific design** docs for details
4. Reference **ADRs** in ../decisions/ for architectural decisions
Custom slugs, author discovery and microformats2, media upload and display.
### For Planning New Features
1. Review similar **phase documents** for patterns
2. Check **project-structure.md** for organization guidelines
3. Create new design doc following existing format
4. Update this index with the new document
- [2025-11-28-v1.2.0-design-complete.md](v1.2.0/2025-11-28-v1.2.0-design-complete.md) — v1.2.0 Design Review - Complete
- [2025-11-28-v1.2.0-media-display-fixes.md](v1.2.0/2025-11-28-v1.2.0-media-display-fixes.md) — Media Display Fixes Implementation Report
- [2025-11-28-v1.2.0-phase1-custom-slugs.md](v1.2.0/2025-11-28-v1.2.0-phase1-custom-slugs.md) — v1.2.0 Phase 1: Custom Slugs - Implementation Report
- [2025-11-28-v1.2.0-phase1-review.md](v1.2.0/2025-11-28-v1.2.0-phase1-review.md) — v1.2.0 Phase 1: Custom Slugs - Architectural Review
- [2025-11-28-v1.2.0-phase2-author-microformats.md](v1.2.0/2025-11-28-v1.2.0-phase2-author-microformats.md) — v1.2.0 Phase 2 Implementation Report: Author Discovery & Microformats2
- [2025-11-28-v1.2.0-phase2-review.md](v1.2.0/2025-11-28-v1.2.0-phase2-review.md) — v1.2.0 Phase 2 Architectural Review: Author Discovery & Microformats2
- [2025-11-28-v1.2.0-phase3-media-upload.md](v1.2.0/2025-11-28-v1.2.0-phase3-media-upload.md) — v1.2.0 Phase 3: Media Upload - Implementation Report
- [2025-11-28-v1.2.0-phase3-review.md](v1.2.0/2025-11-28-v1.2.0-phase3-review.md) — v1.2.0 Phase 3 Architecture Review: Media Upload
- [2025-12-09-feed-media-implementation.md](v1.2.0/2025-12-09-feed-media-implementation.md) — Feed Media Enhancement Implementation Report
- [2025-12-09-media-display-validation.md](v1.2.0/2025-12-09-media-display-validation.md) — Media Display Implementation Validation Report
- [2025-12-09-v1.2.0-release.md](v1.2.0/2025-12-09-v1.2.0-release.md) — v1.2.0 Release Report
- [2025-12-10-documentation-audit-v1.2.0.md](v1.2.0/2025-12-10-documentation-audit-v1.2.0.md) — Documentation Audit Report - Post v1.2.0 Release
- [media-display-fixes.md](v1.2.0/media-display-fixes.md) — Media Display Fixes - Architectural Design
- [v1.2.0-media-css-design.md](v1.2.0/v1.2.0-media-css-design.md) — CSS Design for Media Display (v1.2.0)
- [v1.X.X-indieweb-options.md](v1.2.0/v1.X.X-indieweb-options.md) — StarPunk v1.X.X IndieWeb-Focused Release Options
- [v1.X.X-options.md](v1.2.0/v1.X.X-options.md) — StarPunk Next Release Options
### For Understanding Existing Code
1. Find the **phase** that implemented the feature
2. Read the design document for context
3. Check **ADRs** for decision rationale
4. Review implementation reports in ../reports/
## v1.3.0
## Document Types
Microformats2 compliance and tags.
### Phase Documents
Comprehensive plans for each development phase, including:
- Goals and scope
- Implementation tasks
- Dependencies
- Testing requirements
- [2025-12-10-phase1-implementation.md](v1.3.0/2025-12-10-phase1-implementation.md) — v1.3.0 Phase 1 Implementation Report
- [2025-12-10-phase2-implementation.md](v1.3.0/2025-12-10-phase2-implementation.md) — v1.3.0 Phase 2: Templates Implementation Report
- [2025-12-10-phase3-implementation.md](v1.3.0/2025-12-10-phase3-implementation.md) — v1.3.0 Phase 3 Implementation Report
- [2025-12-10-phase4-implementation.md](v1.3.0/2025-12-10-phase4-implementation.md) — v1.3.0 Phase 4: Validation Implementation Report
- [microformats-tags-design.md](v1.3.0/microformats-tags-design.md) — v1.3.0 Microformats2 Compliance and Tags Implementation Design
### Quick Reference Documents
Condensed information for rapid development:
- Key decisions
- Code patterns
- Common operations
- Gotchas and notes
## v1.3.1
### Feature Design Documents
Detailed specifications for specific features:
- Requirements
- API design
- Data models
- UI/UX considerations
Tags in feeds.
### Diagnostic Documents
Problem analysis and solutions:
- Issue description
- Root cause analysis
- Solution design
- Implementation plan
- [2025-12-10-feed-tags-implementation.md](v1.3.1/2025-12-10-feed-tags-implementation.md) — Feed Tags Implementation Report
- [feed-tags-design.md](v1.3.1/feed-tags-design.md) — Feed Tags Implementation Design
## v1.4.0
"Media" release: image variants and media pipeline.
- [2025-12-16-phase4-implementation-report.md](v1.4.0/2025-12-16-phase4-implementation-report.md) — v1.4.0 Phase 4 Implementation Report
- [2025-12-16-v140-implementation-complete.md](v1.4.0/2025-12-16-v140-implementation-complete.md) — v1.4.0 "Media" Release - Implementation Complete
- [media-implementation-design.md](v1.4.0/media-implementation-design.md) — v1.4.0 "Media" Release - Implementation Design
## v1.4.1
Media upload logging.
- [2025-12-16-implementation-report.md](v1.4.1/2025-12-16-implementation-report.md) — Implementation Report - v1.4.1 Media Upload Logging
- [architect-responses.md](v1.4.1/architect-responses.md) — Architect Responses - v1.4.1 Media Upload Logging
- [developer-questions.md](v1.4.1/developer-questions.md) — Developer Questions - v1.4.1 Media Upload Logging
- [media-logging-design.md](v1.4.1/media-logging-design.md) — Media Upload Logging Design - v1.4.1
## v1.4.2
HEIC/MPO support and larger dimension limit.
- [2025-12-16-architect-review.md](v1.4.2/2025-12-16-architect-review.md) — Architect Review: v1.4.2 Implementation and End-to-End Media Pipeline
- [2025-12-16-developer-review.md](v1.4.2/2025-12-16-developer-review.md) — v1.4.2 Developer Review - HEIC/MPO Support and Dimension Limit Increase
- [2025-12-16-implementation-report.md](v1.4.2/2025-12-16-implementation-report.md) — v1.4.2 Implementation Report - HEIC Image Support
- [heic-support-design.md](v1.4.2/heic-support-design.md) — HEIC Image Support Design - v1.4.2
## v1.5.0
Timestamp slugs, debug file management, N+1 fix, atomic variants, test coverage.
- [2025-12-16-architect-responses.md](v1.5.0/2025-12-16-architect-responses.md) — Architect Responses - v1.5.0 Developer Questions
- [2025-12-16-developer-questions.md](v1.5.0/2025-12-16-developer-questions.md) — Developer Review Questions - v1.5.0 Release Plan
- [2025-12-16-phase-0-implementation-report.md](v1.5.0/2025-12-16-phase-0-implementation-report.md) — Phase 0 Implementation Report - Test Fixes
- [2025-12-16-phase2-architect-review.md](v1.5.0/2025-12-16-phase2-architect-review.md) — Phase 2 Architect Review: Debug File Management
- [2025-12-17-indieauth-discovery-fix.md](v1.5.0/2025-12-17-indieauth-discovery-fix.md) — IndieAuth Endpoint Discovery Regression Analysis
- [2025-12-17-phase-1-implementation-report.md](v1.5.0/2025-12-17-phase-1-implementation-report.md) — Phase 1 Implementation Report: Timestamp-Based Slugs
- [2025-12-17-phase3-architect-review.md](v1.5.0/2025-12-17-phase3-architect-review.md) — Phase 3 Architect Review: N+1 Query Fix
- [2025-12-17-phase3-implementation.md](v1.5.0/2025-12-17-phase3-implementation.md) — v1.5.0 Phase 3 Implementation Report
- [2025-12-17-phase4-architect-review.md](v1.5.0/2025-12-17-phase4-architect-review.md) — Phase 4 Architect Review: Atomic Variant Generation
- [2025-12-17-phase4-implementation.md](v1.5.0/2025-12-17-phase4-implementation.md) — v1.5.0 Phase 4: Atomic Variant Generation - Implementation Report
- [2025-12-17-phase5-architect-review.md](v1.5.0/2025-12-17-phase5-architect-review.md) — Phase 5 Architect Review: Test Coverage Expansion (FINAL)
- [2025-12-17-phase5-implementation.md](v1.5.0/2025-12-17-phase5-implementation.md) — v1.5.0 Phase 5: Test Coverage Expansion - Implementation Report
- [phase-2-implementation-report.md](v1.5.0/phase-2-implementation-report.md) — v1.5.0 Phase 2 Implementation Report: Debug File Management
## v1.6.0
Webmentions (receive, send, moderate), reply contexts, rate limiting. Original documents lost; reconstructed from the code.
- [api-contract.md](v1.6.0/api-contract.md) — v1.6.0 Design: Webmention API Contract *(reconstructed)*
- [background-worker.md](v1.6.0/background-worker.md) — v1.6.0 Design: Webmention Background Worker *(reconstructed)*
- [database-schema.md](v1.6.0/database-schema.md) — v1.6.0 Design: Webmention Database Schema *(reconstructed)*
- [display-moderation.md](v1.6.0/display-moderation.md) — v1.6.0 Design: Webmention Display and Moderation *(reconstructed)*
- [rate-limiting.md](v1.6.0/rate-limiting.md) — v1.6.0 Design: Webmention Rate Limiting *(reconstructed)*
## v2.0.0
OIDC admin login and built-in IndieAuth server.
- [2026-10-31-oidc-migration.md](v2.0.0/2026-10-31-oidc-migration.md) — v2.0.0 Implementation Report: OIDC Login and Built-in IndieAuth Server
## v2.0.1
- [2026-10-31-absolute-urls-and-ci-buildkit.md](v2.0.1/2026-10-31-absolute-urls-and-ci-buildkit.md) — Absolute URLs from SITE_URL; CI builds with BuildKit *(reconstructed)*
## v2.0.2
- [2026-10-31-ci-registry-token.md](v2.0.2/2026-10-31-ci-registry-token.md) — CI registry login with REGISTRY_TOKEN *(reconstructed)*
## v2.1.0
- [2026-10-31-gotosocial-crossposting.md](v2.1.0/2026-10-31-gotosocial-crossposting.md) — Crossposting to GoToSocial *(reconstructed)*
## v2.2.0
- [2026-10-31-bluesky-crossposting.md](v2.2.0/2026-10-31-bluesky-crossposting.md) — Crossposting to Bluesky *(reconstructed)*
## v2.3.0
- [2026-10-31-public-page-restyle.md](v2.3.0/2026-10-31-public-page-restyle.md) — Public page restyle *(reconstructed)*
## v2.3.1
- [2026-10-31-login-link-in-header.md](v2.3.1/2026-10-31-login-link-in-header.md) — "Log in" back in the header *(reconstructed)*
## v2.3.2
- [2026-10-31-header-admin-link.md](v2.3.2/2026-10-31-header-admin-link.md) — Header shows "Admin" when logged in *(reconstructed)*
## v2.3.3
- [2026-10-31-micropub-title-and-source-url.md](v2.3.3/2026-10-31-micropub-title-and-source-url.md) — Micropub titles, q=source URL, single title display
## v2.4.0
- [2026-10-31-micropub-update-delete-drafts.md](v2.4.0/2026-10-31-micropub-update-delete-drafts.md) — Micropub update, delete/undelete, drafts, replies *(reconstructed)*
## v2.5.0
- [2026-10-31-likes-and-reposts.md](v2.5.0/2026-10-31-likes-and-reposts.md) — Likes and reposts *(reconstructed)*
## v2.6.0
- [2026-10-31-threads-crossposting.md](v2.6.0/2026-10-31-threads-crossposting.md) — Crossposting to Threads
## v2.6.1
- [2026-10-31-threads-publish-retry.md](v2.6.1/2026-10-31-threads-publish-retry.md) — Retry Threads publish errors *(reconstructed)*
## v2.6.2
- [2026-10-31-threads-link-preview.md](v2.6.2/2026-10-31-threads-link-preview.md) — Let Threads build the link preview *(reconstructed)*
## v2.6.3
- [2026-11-01-crosspost-links-and-cards.md](v2.6.3/2026-11-01-crosspost-links-and-cards.md) — Crosspost cards show the note's own link *(reconstructed)*
## v2.6.4
- [2026-11-01-tag-entry-order.md](v2.6.4/2026-11-01-tag-entry-order.md) — Tags keep the order they were entered in *(reconstructed)*
## v2.6.5
- [2026-11-01-micropub-media-fixes.md](v2.6.5/2026-11-01-micropub-media-fixes.md) — Micropub media fixes
## v2.6.6
- [2026-11-02-crosspost-embedded-images.md](v2.6.6/2026-11-02-crosspost-embedded-images.md) — Crossposts include photos embedded in the text *(reconstructed)*
## v2.6.7
- [2026-11-02-admin-layout-width.md](v2.6.7/2026-11-02-admin-layout-width.md) — Admin pages line up with the header *(reconstructed)*
## v2.6.8
- [2026-11-03-webmention-worker-app-context.md](v2.6.8/2026-11-03-webmention-worker-app-context.md) — Webmention worker app context
## v2.6.9
- [2026-11-03-webmention-processing-fixes.md](v2.6.9/2026-11-03-webmention-processing-fixes.md) — Webmention processing fixes
## v2.7.0
- [2026-11-03-webmention-spec-compliance.md](v2.7.0/2026-11-03-webmention-spec-compliance.md) — Webmention W3C compliance
## v2.7.1
- [2026-11-03-ci-test-stage.md](v2.7.1/2026-11-03-ci-test-stage.md) — Tests run before release images are built *(reconstructed)*
## v2.7.2
- [2026-11-03-webmention-shoulds.md](v2.7.2/2026-11-03-webmention-shoulds.md) — Remaining Webmention SHOULDs
## v2.8.0
- [2026-11-03-microformats2-review.md](v2.8.0/2026-11-03-microformats2-review.md) — Microformats2 review
## v2.8.1
- [2026-11-03-migration-lock.md](v2.8.1/2026-11-03-migration-lock.md) — Migration lock held for the whole run
## v2.9.0
- [2026-11-03-micropub-review.md](v2.9.0/2026-11-03-micropub-review.md) — Micropub review
## v2.10.0
- [2026-11-03-websub-implementation.md](v2.10.0/2026-11-03-websub-implementation.md) — WebSub with a built-in hub (implementation report) *(reconstructed)*
## v2.10.1
- [2026-11-03-websub-hardening.md](v2.10.1/2026-11-03-websub-hardening.md) — WebSub hardening (implementation report) *(reconstructed)*
## v2.10.2
- [2026-11-03-reply-context-fetch.md](v2.10.2/2026-11-03-reply-context-fetch.md) — Reply context fetch
## v2.11.0
- [2026-11-03-maintenance-release.md](v2.11.0/2026-11-03-maintenance-release.md) — Maintenance release
## v2.12.0
- [2026-11-03-remove-metrics.md](v2.12.0/2026-11-03-remove-metrics.md) — Remove the metrics dashboard
## v2.12.1
- [2026-11-10-category-unused-tags.md](v2.12.1/2026-11-10-category-unused-tags.md) — q=category lists only tags in use
## v2.13.0
- [2026-11-10-backfeed.md](v2.13.0/2026-11-10-backfeed.md) — Backfeed (implementation report)
## v2.13.1
- [2026-11-10-avatar-fetch.md](v2.13.1/2026-11-10-avatar-fetch.md) — Avatar fetching in production
## v2.13.2
- [2026-11-10-reaction-counts.md](v2.13.2/2026-11-10-reaction-counts.md) — Reaction counts and web actions
## v2.13.3
- [2026-11-10-count-opted-out.md](v2.13.3/2026-11-10-count-opted-out.md) — Count opted-out people without naming them
## v2.13.4
- [2026-11-10-list-counts.md](v2.13.4/2026-11-10-list-counts.md) — Counts on lists of notes, reply cards as links
## v2.14.0
- [2026-11-10-notifications.md](v2.14.0/2026-11-10-notifications.md) — Notifications (implementation report)
## v2.14.1
- [2026-11-10-tags-and-captions.md](v2.14.1/2026-11-10-tags-and-captions.md) — Clearing tags in the editor; captions escaped in feeds
## v2.14.2
- [2026-11-10-quotes-cleanup.md](v2.14.2/2026-11-10-quotes-cleanup.md) — Quote counts; dead files removed
## hotfix
Not a release folder.
IndieAuth PKCE endpoint discovery hotfix (2025-12-17, between v1.5.0 and v1.6.0; no version of its own).
- [2025-12-17-indieauth-pkce-endpoint-discovery-hotfix-implementation-report.md](hotfix/2025-12-17-indieauth-pkce-endpoint-discovery-hotfix-implementation-report.md) — IndieAuth Endpoint Discovery Hotfix - Implementation Report
- [2025-12-17-indieauth-pkce-endpoint-discovery-hotfix.md](hotfix/2025-12-17-indieauth-pkce-endpoint-discovery-hotfix.md) — IndieAuth Endpoint Discovery Hotfix
## Related Documentation
- **[../architecture/](../architecture/)** - System architecture and overviews
- **[../decisions/](../decisions/)** - Architectural Decision Records (ADRs)
- **[../reports/](../reports/)** - Implementation reports
- **[../standards/](../standards/)** - Coding standards and conventions
- [../architecture/](../architecture/) - Current system architecture
- [../decisions/](../decisions/) - Architectural Decision Records (ADRs)
- [../projectplan/](../projectplan/) - Backlog and release plans
- [../standards/](../standards/) - Coding standards and conventions
- [../../CHANGELOG.md](../../CHANGELOG.md) - User-facing release notes
---
**Last Updated**: 2025-11-25
**Maintained By**: Documentation Manager Agent
**Last Updated**: 2026-11-10
@@ -0,0 +1,173 @@
# IndieAuth Endpoint Discovery Hotfix - Implementation Report
**Date:** 2025-12-17
**Type:** Production Hotfix
**Priority:** Critical
**Status:** Implementation Complete
## Summary
Successfully implemented the IndieAuth endpoint discovery hotfix as specified in the design document. The authentication flow now correctly discovers endpoints from the user's profile URL per the W3C IndieAuth specification, instead of hardcoding the indielogin.com service.
## Implementation Steps
All implementation steps were completed successfully:
### Step 1: Update starpunk/config.py - Remove INDIELOGIN_URL
- Removed `INDIELOGIN_URL` config line (previously line 37)
- Added deprecation warning for users still setting `INDIELOGIN_URL` in environment
- Warning directs users to remove the deprecated config
### Step 2: Update starpunk/auth.py - Add imports and use endpoint discovery
- Added imports: `discover_endpoints`, `DiscoveryError`, `normalize_url` from `starpunk.auth_external`
- Rewrote `initiate_login()`:
- Now discovers authorization_endpoint from the user's profile URL
- Uses discovered endpoint instead of hardcoded INDIELOGIN_URL
- Raises DiscoveryError if endpoint discovery fails or no authorization_endpoint found
- Rewrote `handle_callback()`:
- Discovers authorization_endpoint from ADMIN_ME profile
- Uses authorization_endpoint for authentication-only flow (per IndieAuth spec)
- Does NOT include `grant_type` parameter (not needed for auth-only flows)
- Uses `normalize_url()` for URL comparison to handle trailing slashes and case differences
### Step 3: Update starpunk/auth_external.py - Relax endpoint validation
- Changed endpoint validation in `_fetch_and_parse()`:
- Now requires at least one endpoint (authorization_endpoint OR token_endpoint)
- Previously required token_endpoint to be present
- This allows profiles with only authorization_endpoint to work for login
- Micropub will still require token_endpoint and fail gracefully with 401
### Step 4: Update starpunk/routes/auth.py - Import and handle DiscoveryError
- Added import for `DiscoveryError` from `starpunk.auth_external`
- Added exception handler in `login_initiate()`:
- Catches DiscoveryError
- Logs technical details at ERROR level
- Shows user-friendly message: "Unable to verify your profile URL. Please check that it's correct and try again."
- Redirects back to login form
### Step 5: Update tests/test_auth.py - Mock discover_endpoints()
- Removed `INDIELOGIN_URL` from test app fixture
- Updated all tests that call `initiate_login()` or `handle_callback()`:
- Added `@patch("starpunk.auth.discover_endpoints")` decorator
- Mock returns both authorization_endpoint and token_endpoint
- Updated assertions to check for discovered endpoint instead of indielogin.com
- Tests updated:
- `TestInitiateLogin.test_initiate_login_success`
- `TestInitiateLogin.test_initiate_login_stores_state`
- `TestHandleCallback.test_handle_callback_success`
- `TestHandleCallback.test_handle_callback_unauthorized_user`
- `TestHandleCallback.test_handle_callback_indielogin_error`
- `TestHandleCallback.test_handle_callback_no_identity`
- `TestLoggingIntegration.test_initiate_login_logs_at_debug`
- `TestLoggingIntegration.test_initiate_login_info_level`
- `TestLoggingIntegration.test_handle_callback_logs_http_details`
### Step 6: Update tests/test_auth_external.py - Fix error message
- Updated `test_discover_endpoints_no_token_endpoint`:
- Changed assertion from "No token endpoint found" to "No IndieAuth endpoints found"
- Matches new relaxed validation error message
### Step 7: Run tests to verify implementation
- All 51 tests in `tests/test_auth.py` pass
- All 35 tests in `tests/test_auth_external.py` pass
- All 32 tests in `tests/test_routes_admin.py` pass
- No regressions detected
## Files Modified
| File | Lines Changed | Description |
|------|--------------|-------------|
| `starpunk/config.py` | 9 added, 1 removed | Removed INDIELOGIN_URL, added deprecation warning |
| `starpunk/auth.py` | 1 added, 84 replaced | Added imports, rewrote initiate_login() and handle_callback() |
| `starpunk/auth_external.py` | 6 replaced | Relaxed endpoint validation |
| `starpunk/routes/auth.py` | 5 added | Added DiscoveryError import and exception handling |
| `tests/test_auth.py` | 1 removed, 43 modified | Removed INDIELOGIN_URL from fixture, added mocks |
| `tests/test_auth_external.py` | 2 modified | Updated error message assertion |
## Key Implementation Details
### Authorization Endpoint Usage
Per IndieAuth spec and architect clarifications:
- Authentication-only flows POST to the **authorization_endpoint** (not token_endpoint)
- The `grant_type` parameter is NOT included (only for access token flows)
- This differs from the previous implementation which incorrectly used indielogin.com's endpoints
### URL Normalization
The implementation now uses `normalize_url()` when comparing the returned 'me' URL with ADMIN_ME:
- Handles trailing slash differences (https://example.com vs https://example.com/)
- Handles case differences (https://Example.com vs https://example.com)
- This is spec-compliant behavior that was previously missing
### Error Handling
- Discovery failures are caught and logged at ERROR level
- User-facing error message is simplified and friendly
- Technical details remain in logs for debugging
### Backward Compatibility
- Deprecation warning added for INDIELOGIN_URL environment variable
- Existing .env files with INDIELOGIN_URL will log a warning but continue to work
- Users are instructed to remove the deprecated config
## Testing Results
### Unit Tests
- `tests/test_auth.py`: 51/51 passed
- `tests/test_auth_external.py`: 35/35 passed
- `tests/test_routes_admin.py`: 32/32 passed
### Integration
All modified tests now correctly mock endpoint discovery and validate the new behavior.
## Issues Encountered
No significant issues encountered during implementation. The design document was thorough and all architect clarifications were addressed:
1. Import placement - Moved to top-level as specified
2. URL normalization - Included as intentional bugfix
3. Endpoint selection - Used authorization_endpoint for auth-only flows
4. Validation relaxation - Allowed profiles with only authorization_endpoint
5. Test strategy - Mocked discover_endpoints() and removed INDIELOGIN_URL
6. grant_type parameter - Correctly omitted for auth-only flows
7. Error messages - Simplified user-facing messages
## Next Steps
1. Manual testing recommended:
- Test login flow with actual IndieAuth profile
- Verify endpoint discovery logs appear
- Test with profiles that have custom endpoints
- Verify error message appears for profiles without endpoints
2. Deployment:
- Update production .env to remove INDIELOGIN_URL (optional - will show warning)
- Deploy changes
- Monitor logs for "Discovered authorization_endpoint" messages
- Verify login works for admin user
3. Documentation:
- Update CHANGELOG.md with hotfix entry
- Consider adding migration guide if needed
## Verification Checklist
- [x] All specified files modified
- [x] All code changes follow architect's design exactly
- [x] Tests updated and passing
- [x] Error messages user-friendly
- [x] Logging appropriate for debugging
- [x] URL normalization implemented
- [x] Endpoint validation relaxed correctly
- [x] No regressions in existing tests
- [x] Implementation report created
## Conclusion
The hotfix has been successfully implemented according to the architect's design. The authentication flow now correctly implements the W3C IndieAuth specification for endpoint discovery. All tests pass and no regressions were detected.
The critical production bug preventing user login should be resolved once this code is deployed.
---
**Developer:** Claude (Fullstack Developer Subagent)
**Date Completed:** 2025-12-17
**Ready for Review:** Yes
@@ -0,0 +1,608 @@
# IndieAuth Endpoint Discovery Hotfix
**Date:** 2025-12-17
**Type:** Production Hotfix
**Priority:** Critical
**Status:** Ready for Implementation
## Problem Summary
Users cannot log in to StarPunk. The root cause is that the authentication code ignores endpoint discovery and hardcodes `INDIELOGIN_URL` instead of discovering the authorization and token endpoints from the user's profile URL.
**Root Cause:** The `starpunk/auth.py` module uses `INDIELOGIN_URL` config instead of discovering endpoints from the user's profile URL as required by the IndieAuth specification. This is a regression - the system used to respect discovered endpoints.
**Note:** The PKCE error message in the callback is a symptom, not the cause. Once we use the correct discovered endpoints, PKCE will not be required (since the user's actual IndieAuth server doesn't require it).
## Specification Requirements
### W3C IndieAuth Spec (https://www.w3.org/TR/indieauth/)
- Clients MUST discover `authorization_endpoint` from user's profile URL
- Clients MUST discover `token_endpoint` from user's profile URL
- Discovery via HTTP Link headers (highest priority) or HTML `<link>` elements
## Implementation Plan
### Overview
The fix reuses the existing `discover_endpoints()` function from `auth_external.py` in the login flow. Changes are minimal and focused:
1. Use `discover_endpoints()` in `initiate_login()` to get the `authorization_endpoint`
2. Use `discover_endpoints()` in `handle_callback()` to get the `token_endpoint`
3. Remove `INDIELOGIN_URL` config (with deprecation warning)
---
### Step 1: Update config.py - Remove INDIELOGIN_URL
In `/home/phil/Projects/starpunk/starpunk/config.py`:
**Change 1:** Remove the INDIELOGIN_URL config line (line 37):
```python
# DELETE THIS LINE:
app.config["INDIELOGIN_URL"] = os.getenv("INDIELOGIN_URL", "https://indielogin.com")
```
**Change 2:** Add deprecation warning for INDIELOGIN_URL (add after the TOKEN_ENDPOINT warning, around line 47):
```python
# DEPRECATED: INDIELOGIN_URL no longer used (hotfix 2025-12-17)
# Authorization endpoint is now discovered from ADMIN_ME profile per IndieAuth spec
if 'INDIELOGIN_URL' in os.environ:
app.logger.warning(
"INDIELOGIN_URL is deprecated and will be ignored. "
"Remove it from your configuration. "
"The authorization endpoint is now discovered automatically from your ADMIN_ME profile."
)
```
---
### Step 2: Update auth.py - Use Endpoint Discovery
In `/home/phil/Projects/starpunk/starpunk/auth.py`:
**Change 1:** Add import for endpoint discovery (after line 42):
```python
from starpunk.auth_external import discover_endpoints, DiscoveryError, normalize_url
```
> **Note:** The `normalize_url` import is at the top level (not inside `handle_callback()`) for consistency with the existing code style.
**Change 2:** Update `initiate_login()` to use discovered authorization_endpoint (replace lines 251-318):
```python
def initiate_login(me_url: str) -> str:
"""
Initiate IndieAuth authentication flow with endpoint discovery.
Per W3C IndieAuth spec, discovers authorization_endpoint from user's
profile URL rather than using a hardcoded service.
Args:
me_url: User's IndieWeb identity URL
Returns:
Redirect URL to discovered authorization endpoint
Raises:
ValueError: Invalid me_url format
DiscoveryError: Failed to discover endpoints from profile
"""
# Validate URL format
if not is_valid_url(me_url):
raise ValueError(f"Invalid URL format: {me_url}")
current_app.logger.debug(f"Auth: Validating me URL: {me_url}")
# Discover authorization endpoint from user's profile URL
# Per IndieAuth spec: clients MUST discover endpoints, not hardcode them
try:
endpoints = discover_endpoints(me_url)
except DiscoveryError as e:
current_app.logger.error(f"Auth: Endpoint discovery failed for {me_url}: {e}")
raise
auth_endpoint = endpoints.get('authorization_endpoint')
if not auth_endpoint:
raise DiscoveryError(
f"No authorization_endpoint found at {me_url}. "
"Ensure your profile has IndieAuth link elements or headers."
)
current_app.logger.info(f"Auth: Discovered authorization_endpoint: {auth_endpoint}")
# Generate CSRF state token
state = _generate_state_token()
current_app.logger.debug(f"Auth: Generated state token: {_redact_token(state, 8)}")
# Store state in database (5-minute expiry)
db = get_db(current_app)
expires_at = datetime.utcnow() + timedelta(minutes=5)
redirect_uri = f"{current_app.config['SITE_URL']}auth/callback"
db.execute(
"""
INSERT INTO auth_state (state, expires_at, redirect_uri)
VALUES (?, ?, ?)
""",
(state, expires_at, redirect_uri),
)
db.commit()
# Build authorization URL
params = {
"me": me_url,
"client_id": current_app.config["SITE_URL"],
"redirect_uri": redirect_uri,
"state": state,
"response_type": "code",
}
current_app.logger.debug(
f"Auth: Building authorization URL with params:\n"
f" me: {me_url}\n"
f" client_id: {current_app.config['SITE_URL']}\n"
f" redirect_uri: {redirect_uri}\n"
f" state: {_redact_token(state, 8)}\n"
f" response_type: code"
)
auth_url = f"{auth_endpoint}?{urlencode(params)}"
current_app.logger.debug(f"Auth: Complete authorization URL: {auth_url}")
current_app.logger.info(f"Auth: Authentication initiated for {me_url}")
return auth_url
```
**Change 3:** Update `handle_callback()` to use discovered authorization_endpoint (replace lines 321-474):
> **Important:** Per IndieAuth spec, authentication-only flows (identity verification without access tokens) POST to the **authorization_endpoint**, NOT the token_endpoint. The `grant_type` parameter is NOT included for authentication-only flows.
```python
def handle_callback(code: str, state: str, iss: Optional[str] = None) -> Optional[str]:
"""
Handle IndieAuth callback with endpoint discovery.
Discovers authorization_endpoint from ADMIN_ME profile and exchanges
authorization code for identity verification.
Per IndieAuth spec: Authentication-only flows POST to the authorization
endpoint (not token endpoint) and do not include grant_type.
Args:
code: Authorization code from authorization server
state: CSRF state token
iss: Issuer identifier (optional, for security validation)
Returns:
Session token if successful, None otherwise
Raises:
InvalidStateError: State token validation failed
UnauthorizedError: User not authorized as admin
IndieLoginError: Code exchange failed
"""
current_app.logger.debug(f"Auth: Verifying state token: {_redact_token(state, 8)}")
# Verify state token (CSRF protection)
if not _verify_state_token(state):
current_app.logger.warning(
"Auth: Invalid state token received (possible CSRF or expired token)"
)
raise InvalidStateError("Invalid or expired state token")
current_app.logger.debug("Auth: State token valid")
# Discover authorization endpoint from ADMIN_ME profile
admin_me = current_app.config.get("ADMIN_ME")
if not admin_me:
current_app.logger.error("Auth: ADMIN_ME not configured")
raise IndieLoginError("ADMIN_ME not configured")
try:
endpoints = discover_endpoints(admin_me)
except DiscoveryError as e:
current_app.logger.error(f"Auth: Endpoint discovery failed: {e}")
raise IndieLoginError(f"Failed to discover endpoints: {e}")
# Use authorization_endpoint for authentication-only flow (identity verification)
# Per IndieAuth spec: auth-only flows POST to authorization_endpoint, not token_endpoint
auth_endpoint = endpoints.get('authorization_endpoint')
if not auth_endpoint:
raise IndieLoginError(
f"No authorization_endpoint found at {admin_me}. "
"Ensure your profile has IndieAuth endpoints configured."
)
current_app.logger.debug(f"Auth: Using authorization_endpoint: {auth_endpoint}")
# Verify issuer if provided (security check - optional)
if iss:
current_app.logger.debug(f"Auth: Issuer provided: {iss}")
# Prepare code verification request
# Note: grant_type is NOT included for authentication-only flows per IndieAuth spec
token_exchange_data = {
"code": code,
"client_id": current_app.config["SITE_URL"],
"redirect_uri": f"{current_app.config['SITE_URL']}auth/callback",
}
# Log the request
_log_http_request(
method="POST",
url=auth_endpoint,
data=token_exchange_data,
)
current_app.logger.debug(
"Auth: Sending code verification request to authorization endpoint:\n"
" Method: POST\n"
" URL: %s\n"
" Data: code=%s, client_id=%s, redirect_uri=%s",
auth_endpoint,
_redact_token(code),
token_exchange_data["client_id"],
token_exchange_data["redirect_uri"],
)
# Exchange code for identity at authorization endpoint (authentication-only flow)
try:
response = httpx.post(
auth_endpoint,
data=token_exchange_data,
timeout=10.0,
)
current_app.logger.debug(
"Auth: Received code verification response:\n"
" Status: %d\n"
" Headers: %s\n"
" Body: %s",
response.status_code,
{k: v for k, v in dict(response.headers).items()
if k.lower() not in ["set-cookie", "authorization"]},
_redact_token(response.text) if response.text else "(empty)",
)
_log_http_response(
status_code=response.status_code,
headers=dict(response.headers),
body=response.text,
)
response.raise_for_status()
except httpx.RequestError as e:
current_app.logger.error(f"Auth: Authorization endpoint request failed: {e}")
raise IndieLoginError(f"Failed to verify code: {e}")
except httpx.HTTPStatusError as e:
current_app.logger.error(
f"Auth: Authorization endpoint returned error: {e.response.status_code} - {e.response.text}"
)
raise IndieLoginError(
f"Authorization endpoint returned error: {e.response.status_code}"
)
# Parse response
try:
data = response.json()
except Exception as e:
current_app.logger.error(f"Auth: Failed to parse authorization endpoint response: {e}")
raise IndieLoginError("Invalid JSON response from authorization endpoint")
me = data.get("me")
if not me:
current_app.logger.error("Auth: No identity returned from authorization endpoint")
raise IndieLoginError("No identity returned from authorization endpoint")
current_app.logger.debug(f"Auth: Received identity: {me}")
# Verify this is the admin user
current_app.logger.info(f"Auth: Verifying admin authorization for me={me}")
# Normalize URLs for comparison (handles trailing slashes and case differences)
# This is correct per IndieAuth spec - the returned 'me' is the canonical form
if normalize_url(me) != normalize_url(admin_me):
current_app.logger.warning(
f"Auth: Unauthorized login attempt: {me} (expected {admin_me})"
)
raise UnauthorizedError(f"User {me} is not authorized")
current_app.logger.debug("Auth: Admin verification passed")
# Create session
session_token = create_session(me)
# Trigger author profile discovery (v1.2.0 Phase 2)
# Per Q14: Never block login, always allow fallback
try:
from starpunk.author_discovery import get_author_profile
author_profile = get_author_profile(me, refresh=True)
current_app.logger.info(f"Author profile refreshed for {me}")
except Exception as e:
current_app.logger.warning(f"Author discovery failed: {e}")
# Continue login anyway - never block per Q14
return session_token
```
---
### Step 3: Update auth_external.py - Relax Endpoint Validation
The existing `_fetch_and_parse()` function requires `token_endpoint` to be present. We need to relax this since some profiles may only have `authorization_endpoint` (for authentication-only flows).
In `/home/phil/Projects/starpunk/starpunk/auth_external.py`, update the validation in `_fetch_and_parse()` (around lines 302-307):
**Change:** Make token_endpoint not strictly required (allow authentication-only profiles):
```python
# Validate we found at least one endpoint
# - authorization_endpoint: Required for authentication-only flows (admin login)
# - token_endpoint: Required for Micropub token verification
# Having at least one allows the appropriate flow to work
if 'token_endpoint' not in endpoints and 'authorization_endpoint' not in endpoints:
raise DiscoveryError(
f"No IndieAuth endpoints found at {profile_url}. "
"Ensure your profile has authorization_endpoint or token_endpoint configured."
)
```
---
### Step 4: Update routes/auth.py - Handle DiscoveryError
In `/home/phil/Projects/starpunk/starpunk/routes/auth.py`:
**Change 1:** Add import for DiscoveryError (update lines 20-29):
```python
from starpunk.auth import (
IndieLoginError,
InvalidStateError,
UnauthorizedError,
destroy_session,
handle_callback,
initiate_login,
require_auth,
verify_session,
)
from starpunk.auth_external import DiscoveryError
```
**Change 2:** Handle DiscoveryError in login_initiate() (update lines 79-85):
> **Note:** The user-facing error message is kept simple. Technical details are logged but not shown to users.
```python
try:
# Initiate IndieAuth flow
auth_url = initiate_login(me_url)
return redirect(auth_url)
except ValueError as e:
flash(str(e), "error")
return redirect(url_for("auth.login_form"))
except DiscoveryError as e:
current_app.logger.error(f"Endpoint discovery failed for {me_url}: {e}")
flash("Unable to verify your profile URL. Please check that it's correct and try again.", "error")
return redirect(url_for("auth.login_form"))
```
---
## File Summary
| File | Change Type | Description |
|------|-------------|-------------|
| `starpunk/config.py` | Edit | Remove INDIELOGIN_URL, add deprecation warning |
| `starpunk/auth.py` | Edit | Use endpoint discovery instead of hardcoded URL |
| `starpunk/auth_external.py` | Edit | Relax endpoint validation (allow auth-only flow) |
| `starpunk/routes/auth.py` | Edit | Handle DiscoveryError exception |
---
## Testing Requirements
### Manual Testing
1. **Login Flow Test**
- Navigate to `/auth/login`
- Enter ADMIN_ME URL
- Verify redirect goes to discovered authorization_endpoint (not hardcoded indielogin.com)
- Complete login and verify session is created
2. **Endpoint Discovery Test**
- Test with profile that declares custom endpoints
- Verify discovered endpoints are used, not defaults
### Existing Test Updates
**Update test fixture in `tests/test_auth.py`:**
Remove `INDIELOGIN_URL` from the app fixture (line 51):
```python
@pytest.fixture
def app(tmp_path):
"""Create Flask app for testing"""
from starpunk import create_app
test_data_dir = tmp_path / "data"
test_data_dir.mkdir(parents=True, exist_ok=True)
app = create_app(
{
"TESTING": True,
"SITE_URL": "http://localhost:5000/",
"ADMIN_ME": "https://example.com",
"SESSION_SECRET": secrets.token_hex(32),
"SESSION_LIFETIME": 30,
# REMOVED: "INDIELOGIN_URL": "https://indielogin.com",
"DATA_PATH": test_data_dir,
"NOTES_PATH": test_data_dir / "notes",
"DATABASE_PATH": test_data_dir / "starpunk.db",
}
)
return app
```
**Update existing tests that use `httpx.post` mock:**
Tests in `TestInitiateLogin` and `TestHandleCallback` need to mock `discover_endpoints()` in addition to `httpx.post`. Example pattern:
```python
@patch("starpunk.auth.discover_endpoints")
@patch("starpunk.auth.httpx.post")
def test_handle_callback_success(self, mock_post, mock_discover, app, db, client):
"""Test successful callback handling"""
# Mock endpoint discovery
mock_discover.return_value = {
'authorization_endpoint': 'https://auth.example.com/authorize',
'token_endpoint': 'https://auth.example.com/token'
}
# Rest of test remains the same...
```
**Update `TestInitiateLogin.test_initiate_login_success`:**
The assertion checking for `indielogin.com` needs to change to check for the mocked endpoint:
```python
@patch("starpunk.auth.discover_endpoints")
def test_initiate_login_success(self, mock_discover, app, db):
"""Test successful login initiation"""
mock_discover.return_value = {
'authorization_endpoint': 'https://auth.example.com/authorize',
'token_endpoint': 'https://auth.example.com/token'
}
with app.app_context():
me_url = "https://example.com"
auth_url = initiate_login(me_url)
# Changed: Check for discovered endpoint instead of indielogin.com
assert "auth.example.com/authorize" in auth_url
assert "me=https%3A%2F%2Fexample.com" in auth_url
# ... rest of assertions
```
### New Automated Tests to Add
```python
# tests/test_auth_endpoint_discovery.py
def test_initiate_login_uses_endpoint_discovery(client, mocker):
"""Verify login uses discovered endpoint, not hardcoded"""
mock_discover = mocker.patch('starpunk.auth.discover_endpoints')
mock_discover.return_value = {
'authorization_endpoint': 'https://custom-auth.example.com/authorize',
'token_endpoint': 'https://custom-auth.example.com/token'
}
response = client.post('/auth/login', data={'me': 'https://example.com'})
assert response.status_code == 302
assert 'custom-auth.example.com' in response.headers['Location']
def test_callback_uses_discovered_authorization_endpoint(client, mocker):
"""Verify callback uses discovered authorization endpoint (not token endpoint)"""
mock_discover = mocker.patch('starpunk.auth.discover_endpoints')
mock_discover.return_value = {
'authorization_endpoint': 'https://custom-auth.example.com/authorize',
'token_endpoint': 'https://custom-auth.example.com/token'
}
mock_post = mocker.patch('starpunk.auth.httpx.post')
# Setup state token and mock httpx response
# Verify code exchange POSTs to authorization_endpoint, not token_endpoint
pass
def test_discovery_error_shows_user_friendly_message(client, mocker):
"""Verify discovery failures show helpful error"""
mock_discover = mocker.patch('starpunk.auth.discover_endpoints')
mock_discover.side_effect = DiscoveryError("No endpoints found")
response = client.post('/auth/login', data={'me': 'https://example.com'})
assert response.status_code == 302
# Should redirect back to login form with flash message
def test_url_normalization_handles_trailing_slash(app, mocker):
"""Verify URL normalization allows trailing slash differences"""
# ADMIN_ME without trailing slash, auth server returns with trailing slash
# Should still authenticate successfully
pass
def test_url_normalization_handles_case_differences(app, mocker):
"""Verify URL normalization is case-insensitive"""
# ADMIN_ME: https://Example.com, auth server returns: https://example.com
# Should still authenticate successfully
pass
```
---
## Rollback Plan
If issues occur after deployment:
1. **Code:** Revert to previous commit
2. **Config:** Re-add INDIELOGIN_URL to .env if needed
---
## Post-Deployment Verification
1. Verify login works with the user's actual profile URL
2. Check logs for "Discovered authorization_endpoint" message
3. Test logout and re-login cycle
---
## Architect Q&A (2025-12-17)
Developer questions answered by the architect prior to implementation:
### Q1: Import Placement
**Q:** Should `normalize_url` import be inside the function or at top level?
**A:** Move to top level with other imports for consistency. The design has been updated.
### Q2: URL Normalization Behavior Change
**Q:** Is the URL normalization change intentional?
**A:** Yes, this is an intentional bugfix. The current exact-match behavior is incorrect per IndieAuth spec. URLs differing only in trailing slashes or case should be considered equivalent for identity purposes. The `normalize_url()` function already exists in `auth_external.py` and is used by `verify_external_token()`.
### Q3: Which Endpoint for Authentication Flow?
**Q:** Should we use token_endpoint or authorization_endpoint?
**A:** Use **authorization_endpoint** for authentication-only flows. Per IndieAuth spec: "the client makes a POST request to the authorization endpoint to verify the authorization code and retrieve the final user profile URL." The design has been corrected.
### Q4: Endpoint Validation Relaxation
**Q:** Is relaxed endpoint validation acceptable?
**A:** Yes. Login requires `authorization_endpoint`, Micropub requires `token_endpoint`. Requiring at least one is correct. If only auth endpoint exists, login works but Micropub fails gracefully (401).
### Q5: Test Update Strategy
**Q:** Remove INDIELOGIN_URL and/or mock discover_endpoints()?
**A:** Both. Remove `INDIELOGIN_URL` from fixtures, add `discover_endpoints()` mocking to existing tests. Detailed guidance added to Testing Requirements section.
### Q6: grant_type Parameter
**Q:** Should we include grant_type in the code exchange?
**A:** No. Authentication-only flows do not include `grant_type`. This parameter is only required when POSTing to the token_endpoint for access tokens. The design has been corrected.
### Q7: Error Message Verbosity
**Q:** Should we simplify the user-facing error message?
**A:** Yes. User-facing message should be simple: "Unable to verify your profile URL. Please check that it's correct and try again." Technical details are logged at ERROR level. The design has been updated.
---
## References
- W3C IndieAuth Specification: https://www.w3.org/TR/indieauth/
- IndieAuth Endpoint Discovery: https://www.w3.org/TR/indieauth/#discovery-by-clients
@@ -1,5 +1,10 @@
# StarPunk Technology Stack - Quick Reference
> **Historical document.** This describes StarPunk's original v0.x technology
> choices (for example IndieLogin.com for admin login) and is kept for
> context only. It does not describe the current system; see
> [technology-stack.md](../../architecture/technology-stack.md) and [overview.md](../../architecture/overview.md).
## Project Understanding
StarPunk is a **minimal, single-user IndieWeb CMS** for publishing notes with RSS syndication. The core philosophy is radical simplicity: "Every line of code must justify its existence."
Loaded 100 of 593 files, more files were not shown because too many files have changed in this diff. Show more