Compare commits

..
6 Commits
Author SHA1 Message Date
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(ADR-073): age-based schedule to 90 days, pruning of withdrawn responses,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
27 changed files with 2044 additions and 41 deletions

No files matched your search

+10 -2
View File
@@ -165,7 +165,8 @@ WEBSUB_ENABLED=true
# 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)
# 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)
@@ -185,7 +186,8 @@ BLUESKY_APP_PASSWORD=
# BLUESKY_MAX_CHARS=300
# Threads (v2.6.0, ADR-069). A long-lived user token with threads_basic,
# threads_content_publish and threads_delete; StarPunk refreshes it daily and
# 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=
@@ -198,6 +200,12 @@ THREADS_USERNAME=
# 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.
# BACKFEED_ENABLED=false
# Stop checking a note's responses after this many days (default: 90)
# BACKFEED_MAX_AGE_DAYS=90
# =============================================================================
# DEBUGGING
# =============================================================================
+21
View File
@@ -7,6 +7,27 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
## [Unreleased]
## [2.13.0] - 2026-11-10
Backfeed: responses to crossposts come back to the note (ADR-073).
### Added
- Likes, reposts, quotes and replies to a note's crossposts on GoToSocial and Bluesky, and replies on Threads, are shown on the note as responses, labelled "via Bluesky" (or GoToSocial, Threads). StarPunk checks each crosspost itself: every 15 minutes on a note's first day, hourly in its first week, daily until 30 days, weekly until `BACKFEED_MAX_AGE_DAYS` (90). Withdrawn likes and deleted replies drop off at the next check
- Threads like, repost and quote counts, shown as a line such as "On Threads: 12 likes · 3 reposts" (the Threads API doesn't say who)
- A Quotes group in a note's responses
- **Backfill responses** on the admin Webmentions page checks every crosspost once, however old the note; each note's edit page shows when its responses were last checked, with **Check now**
- The admin Webmentions list can be filtered by where a response came from, and shows deleted ones under **Deleted**
- Settings `BACKFEED_ENABLED` (default off) and `BACKFEED_MAX_AGE_DAYS`. Backfeed needs read access: a GoToSocial token with `read:statuses` and `read:accounts` (or `read`), and a Threads token with `threads_read_replies` and `threads_manage_insights`. Bluesky app passwords already work
### Changed
- Response authors' avatars are saved on this site (96 px WebP in `DATA_PATH/avatars`) and shown from there, instead of being loaded from other sites. Until a copy is saved the initial is shown
- Deleting a webmention in the admin keeps it as a deleted record, so a re-sent webmention or the next check doesn't bring it back
- People with `#nobridge` or `#nobot` in their profile, and Bluesky accounts that ask not to be shown to logged-out visitors, are not backfed; only public responses are
- A webmention for a response that was already backfed (for example a reply also sent by Bridgy) is ignored, so it isn't shown twice
- The Threads client uses `graph.threads.com`, the host in Meta's current documentation
## [2.12.1] - 2026-11-10
### Fixed
@@ -223,6 +223,49 @@ 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).
## Backfeed (v2.13.0, ADR-073)
`starpunk/backfeed.py` brings responses to the copies back to the note.
After each crosspost pass, the worker loop calls `backfeed.process_due()`,
which 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
skipped and their rows deleted. If the note's shown responses changed,
its webmentions are re-sent (spec 3.1.4). Threads counts go in
`syndications.likes_count`, `reposts_count`, `quotes_count`.
- **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 Webmentions page says which
scopes are needed. 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
+260
View File
@@ -0,0 +1,260 @@
# ADR-073: Backfeed of responses to crossposts
## Status
Accepted (v2.13.0, 2026-11-10)
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.
## 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.
+4 -3
View File
@@ -5,8 +5,8 @@ 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.
**Total ADR files**: 68 (67 numbers: ADR-001 to ADR-044 and ADR-050 to
ADR-072, with two files numbered ADR-012). See [Numbering](#numbering).
**Total ADR files**: 69 (68 numbers: ADR-001 to ADR-044 and ADR-050 to
ADR-073, with two files numbered ADR-012). See [Numbering](#numbering).
Status values: **Accepted** (in force), **Superseded** (replaced by a later
ADR), **Deprecated** (withdrawn, nothing replaces it; none at present),
@@ -85,6 +85,7 @@ See also [IndieWeb protocols](../architecture/indieweb-protocols.md).
| [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). |
### Authentication (current)
@@ -167,7 +168,7 @@ was carried out and is not itself superseded.
## How to create a new ADR
1. Use the next number after the highest existing one (currently ADR-073).
1. Use the next number after the highest existing one (currently ADR-074).
2. Name it `ADR-NNN-brief-descriptive-title.md`.
3. Follow the format of recent ADRs (e.g. ADR-070, ADR-071):
+5 -1
View File
@@ -21,7 +21,7 @@ is in `../architecture/`; decisions are in `../decisions/` (ADRs).
## Contents
[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) · [hotfix](#hotfix)
[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) · [hotfix](#hotfix)
## v1.0.0
@@ -422,6 +422,10 @@ OIDC admin login and built-in IndieAuth server.
- [2026-11-10-category-unused-tags.md](v2.12.1/2026-11-10-category-unused-tags.md) - q=category lists only tags on posts that exist
## v2.13.0
- [2026-11-10-backfeed.md](v2.13.0/2026-11-10-backfeed.md) - Backfeed of responses to crossposts (ADR-073), avatar cache, delete tombstones
## hotfix
Not a release folder.
@@ -0,0 +1,53 @@
# v2.13.0: Backfeed (implementation report)
Implements ADR-073.
## What changed
- `migrations/020_backfeed.sql`: `webmentions` rebuilt with `origin`,
`remote_id` and the `quote` type (rows copied, indexes recreated, plus
`(target_note_id, origin)`); `syndications` gains the `backfeed_*`
columns and Threads counts.
- `starpunk/backfeed.py` (new): `Response`/`Harvest`, schedule, lease,
storing and pruning, opt-outs, backfill, `check_now`, `status`.
- `starpunk/syndication.py`: `responses()` on `MastodonClient`,
`BlueskyClient` and `ThreadsClient`; HTTP status kept on errors
(`exc.status`) so refused credentials can be told apart; Mastodon `_call`
accepts absolute URLs (pagination links); the worker loop runs backfeed
after crossposts; Threads host is `graph.threads.com`.
- `starpunk/avatars.py` (new) and `/avatars/<name>`: local avatar copies
for every response; `local_avatar` template filter.
- Webmentions: Quotes group and "via" label; admin delete is a tombstone
(`moderation_status = 'deleted'`, hidden from the list unless filtered);
origin filter; a webmention can't reset a backfed row, and one whose
source is an already backfed response is rejected ("Already backfed").
- Admin: Backfeed panel with **Backfill responses** on the Webmentions
page; **Check now** and last-checked time per crosspost on the edit page.
- Settings `BACKFEED_ENABLED` (default false), `BACKFEED_MAX_AGE_DAYS` (90).
## Verification
- `tests/test_backfeed.py` (22 tests): schedule; text/facet HTML; storing,
updating and moderation kept; pruning only on complete harvests and
restoring; opt-outs; tombstones; backfill and its pacing; likes/reposts
and drafts skipped; refused credentials; backoff; note page (via label,
quotes, Threads counts, local avatars); avatar save/serve/cleanup;
webmention can't overwrite a backfed row; the three clients against fake
APIs; migration 020 on a pre-020 database keeps rows and foreign keys;
the admin Backfeed panel, backfill and Check now; Threads avatar URLs
(signed, changing per request) aren't refetched while a copy exists.
- Tests for the touched areas (webmentions, syndication, microformats,
migrations, templates, routes): all pass. The full suite runs in CI.
- `test_microformats_v280` expects response authors' `u-photo`: kept via
`<data class="u-photo">` until the avatar is saved.
## Deploying
1. Issue a new GoToSocial token with `read:statuses read:accounts` added
to the current scopes, and re-authorise the Threads app with
`threads_read_replies` and `threads_manage_insights`. Put both in the
Ansible vault.
2. Set `BACKFEED_ENABLED=true` in the StarPunk environment and deploy
(migration 020 runs at start: back up the database first).
3. Run **Backfill responses** once from Admin → Webmentions. Turn Bridgy
off for the Mastodon and Bluesky accounts.
-5
View File
@@ -74,11 +74,6 @@ None open.
- **Where**: `templates/base.html`
- **Source**: Documentation refresh 2026-11-03
### Backfeed responses to crossposts
- **Description**: Idea: bring likes, boosts and replies to StarPunk's own crossposts back to the note as responses. Two options: poll GoToSocial (the token would need a `read` scope), or rely on Bridgy, which sends webmentions. Since v2.8.0, StarPunk ignores webmentions *from* a note's own crossposts, so that rule would need to change to work with Bridgy.
- **Where**: `starpunk/syndication.py`, `starpunk/webmention_worker.py`
- **Source**: v2.8.0 microformats2 review discussion
### Tag-filtered feeds
- **Description**: Tag pages are h-feeds and WebSub topics, but there are no RSS, Atom or JSON feeds per tag (for example `/tag/python/feed.atom`). Tag pages also have no pagination, and the editor has no tag autocomplete.
- **Where**: `starpunk/routes/public.py`, `starpunk/feeds/`, `templates/admin/`
+1 -1
View File
@@ -4,7 +4,7 @@
**Last Updated**: 2026-11-03 (v2.10.1)
StarPunk uses [Semantic Versioning 2.0.0](https://semver.org/):
`MAJOR.MINOR.PATCH`, currently **2.12.1**. Every release is a git tag
`MAJOR.MINOR.PATCH`, currently **2.13.0**. Every release is a git tag
`vX.Y.Z` and a `CHANGELOG.md` entry. Where the number is stored and how to
write the changelog entry is in
[version-implementation-guide.md](version-implementation-guide.md); the
+66
View File
@@ -0,0 +1,66 @@
-- Migration 020: Backfeed of responses to crossposts (v2.13.0, ADR-073)
--
-- webmentions: where a response came from (origin, remote_id) and a new
-- 'quote' type. SQLite can't change a CHECK constraint, so the table is
-- rebuilt (as in migration 003). Nothing references webmentions.
--
-- syndications: when each copy's responses were last checked and are next
-- due, a lease so one gunicorn worker checks a copy, and Threads counts.
CREATE TABLE webmentions_new (
id INTEGER PRIMARY KEY AUTOINCREMENT,
source_url TEXT NOT NULL,
target_url TEXT NOT NULL,
target_note_id INTEGER,
mention_type TEXT NOT NULL DEFAULT 'mention'
CHECK (mention_type IN ('reply', 'like', 'repost', 'bookmark', 'mention', 'quote')),
author_name TEXT,
author_url TEXT,
author_photo TEXT,
content_snippet TEXT,
validation_status TEXT NOT NULL DEFAULT 'pending'
CHECK (validation_status IN ('pending', 'valid', 'invalid', 'error')),
moderation_status TEXT NOT NULL DEFAULT 'pending'
CHECK (moderation_status IN ('pending', 'approved', 'hidden', 'deleted')),
validation_error TEXT,
validated_at TIMESTAMP,
received_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
source_post_url TEXT,
source_published TIMESTAMP,
origin TEXT NOT NULL DEFAULT 'webmention',
remote_id TEXT,
FOREIGN KEY (target_note_id) REFERENCES notes(id) ON DELETE SET NULL,
UNIQUE(source_url, target_url)
);
INSERT INTO webmentions_new (
id, source_url, target_url, target_note_id, mention_type, author_name, author_url,
author_photo, content_snippet, validation_status, moderation_status, validation_error,
validated_at, received_at, updated_at, source_post_url, source_published
)
SELECT
id, source_url, target_url, target_note_id, mention_type, author_name, author_url,
author_photo, content_snippet, validation_status, moderation_status, validation_error,
validated_at, received_at, updated_at, source_post_url, source_published
FROM webmentions;
DROP TABLE webmentions;
ALTER TABLE webmentions_new RENAME TO webmentions;
CREATE INDEX IF NOT EXISTS idx_webmentions_target_note ON webmentions(target_note_id);
CREATE INDEX IF NOT EXISTS idx_webmentions_moderation ON webmentions(moderation_status);
CREATE INDEX IF NOT EXISTS idx_webmentions_validation ON webmentions(validation_status);
CREATE INDEX IF NOT EXISTS idx_webmentions_received ON webmentions(received_at DESC);
CREATE INDEX IF NOT EXISTS idx_webmentions_note_origin ON webmentions(target_note_id, origin);
ALTER TABLE syndications ADD COLUMN backfeed_next_at TIMESTAMP;
ALTER TABLE syndications ADD COLUMN backfeed_checked_at TIMESTAMP;
ALTER TABLE syndications ADD COLUMN backfeed_claimed_until TIMESTAMP;
ALTER TABLE syndications ADD COLUMN backfeed_attempts INTEGER NOT NULL DEFAULT 0;
ALTER TABLE syndications ADD COLUMN backfeed_error TEXT;
ALTER TABLE syndications ADD COLUMN backfeed_backfill INTEGER NOT NULL DEFAULT 0;
ALTER TABLE syndications ADD COLUMN likes_count INTEGER;
ALTER TABLE syndications ADD COLUMN reposts_count INTEGER;
ALTER TABLE syndications ADD COLUMN quotes_count INTEGER;
+10 -2
View File
@@ -217,6 +217,14 @@ def create_app(config=None):
from starpunk.utils import iso_utc
app.add_template_filter(iso_utc, 'iso_utc')
# Template filter - local copy of a response author's avatar, or None (v2.13.0)
@app.template_filter('local_avatar')
def local_avatar_filter(url):
from flask import url_for
from starpunk.avatars import cached_name
name = cached_name(app.config.get('DATA_PATH', 'data'), url)
return url_for('public.avatar', name=name) if name else None
# Template filter - URL to color for facepile (v1.6.0)
@app.template_filter('url_to_color')
def url_to_color_filter(url):
@@ -457,5 +465,5 @@ def create_app(config=None):
# Package version (Semantic Versioning 2.0.0)
# See docs/standards/versioning-strategy.md for details
__version__ = "2.12.1"
__version_info__ = (2, 12, 1)
__version__ = "2.13.0"
__version_info__ = (2, 13, 0)
+144
View File
@@ -0,0 +1,144 @@
"""
Local copies of response authors' avatars (v2.13.0, ADR-073 Decision 6)
Pages used to show other sites' avatar URLs directly, which tells those
sites who is reading, and Bluesky and Threads avatar URLs change or expire.
Each avatar is fetched once (SSRF-safe), shrunk to a 96 px WebP and saved
as DATA_PATH/avatars/<sha256 of its URL>.webp. Pages show the saved copy,
or the coloured initial until there is one.
Paths and URLs are passed in, so the webmention worker can use it. Fetching
needs an app context (the SSRF-safe fetcher reads DEV_MODE).
"""
import hashlib
import io
import logging
import re
import sqlite3
import time
from pathlib import Path
from typing import Optional
logger = logging.getLogger(__name__)
SIZE = 96
MAX_BYTES = 1024 * 1024
NAME_RE = re.compile(r"^[0-9a-f]{64}\.webp$")
CLEANUP_EVERY = 7 * 24 * 3600
def name_for(url: str) -> str:
return hashlib.sha256(url.encode()).hexdigest() + ".webp"
def folder(data_path) -> Path:
return Path(data_path) / "avatars"
def cached_name(data_path, url: Optional[str]) -> Optional[str]:
"""File name of the saved copy of this avatar, if there is one"""
if not url or not url.startswith(("http://", "https://")):
return None
name = name_for(url)
return name if (folder(data_path) / name).is_file() else None
def fetch(data_path, url: Optional[str], site_url: str) -> bool:
"""
Save a copy of an avatar unless there is one. Returns True if one exists
afterwards. Failures are logged and the page keeps showing the initial.
"""
if not url or not url.startswith(("http://", "https://")):
return False
path = folder(data_path) / name_for(url)
if path.is_file():
return True
from PIL import Image
from starpunk.webmention_ssrf import safe_get, user_agent
try:
result = safe_get(url, headers={"User-Agent": user_agent(site_url)},
max_bytes=MAX_BYTES, max_redirects=5, timeout=10.0, truncate=False)
if result.status_code >= 400:
raise ValueError(f"HTTP {result.status_code}")
with Image.open(io.BytesIO(result.content)) as image:
image.load()
image = image.convert("RGBA" if "A" in image.getbands() else "RGB")
side = min(image.size)
left, top = (image.width - side) // 2, (image.height - side) // 2
image = image.crop((left, top, left + side, top + side))
image = image.resize((SIZE, SIZE), Image.LANCZOS)
path.parent.mkdir(parents=True, exist_ok=True)
tmp = path.with_suffix(".tmp")
image.save(tmp, "WEBP", quality=80)
tmp.replace(path)
return True
except Exception as e:
logger.info(f"Could not save avatar {url}: {type(e).__name__}: {e}")
return False
def fill_missing(db_path, data_path, site_url: str, limit: int = 20) -> int:
"""Save avatars of shown responses that have no copy yet. Returns how many were tried."""
conn = sqlite3.connect(db_path)
try:
urls = [row[0] for row in conn.execute(
"""
SELECT DISTINCT author_photo FROM webmentions
WHERE author_photo IS NOT NULL AND author_photo != ''
AND validation_status = 'valid' AND moderation_status = 'approved'
ORDER BY id DESC
"""
)]
finally:
conn.close()
tried = 0
for url in urls:
if tried >= limit:
break
if cached_name(data_path, url) or not url.startswith(("http://", "https://")):
continue
fetch(data_path, url, site_url)
tried += 1
return tried
def remove_unused(db_path, data_path) -> int:
"""Delete saved avatars no stored response uses. Returns how many were removed."""
directory = folder(data_path)
if not directory.is_dir():
return 0
conn = sqlite3.connect(db_path)
try:
used = {name_for(row[0]) for row in conn.execute(
"SELECT DISTINCT author_photo FROM webmentions WHERE author_photo IS NOT NULL")}
finally:
conn.close()
removed = 0
for path in directory.iterdir():
if NAME_RE.match(path.name) and path.name not in used:
path.unlink(missing_ok=True)
removed += 1
return removed
class Upkeep:
"""Runs fill_missing every `fill_every` seconds and remove_unused weekly"""
def __init__(self, fill_every: int = 600):
self.fill_every = fill_every
self._filled_at = 0.0
self._cleaned_at = 0.0
def run(self, db_path, data_path, site_url: str) -> None:
now = time.time()
if now - self._filled_at >= self.fill_every:
self._filled_at = now
fill_missing(db_path, data_path, site_url)
if now - self._cleaned_at >= CLEANUP_EVERY:
self._cleaned_at = now
removed = remove_unused(db_path, data_path)
if removed:
logger.info(f"Removed {removed} unused avatars")
+456
View File
@@ -0,0 +1,456 @@
"""
Backfeed of responses to crossposts (v2.13.0, ADR-073)
StarPunk asks each silo for the responses to its own copies of a note
(likes, reposts, quotes and replies) and stores them in `webmentions`, so
moderation, the Responses section and re-sending webmentions when a note's
responses change (spec 3.1.4) work as they do for received webmentions.
Each synced copy of a published note is checked on a schedule that slows as
the note ages, and not at all after BACKFEED_MAX_AGE_DAYS. A backfill
checks every copy once, whatever its age. Checks run in the syndication
worker's loop (`process_due`), one copy at a time under a lease.
A check that read the whole list marks responses that are no longer there
as invalid, so withdrawn likes and deleted replies drop off the page.
"""
import html
import random
import re
from dataclasses import dataclass, field
from datetime import datetime, timezone
from typing import Dict, List, Optional
from flask import current_app
from starpunk.database import get_db
DAY = 24 * 3600
SCHEDULE = [ # (note younger than, check every), in seconds
(DAY, 15 * 60),
(7 * DAY, 3600),
(30 * DAY, DAY),
]
LATE_INTERVAL = 7 * DAY # from 30 days until the cutoff
CLAIM_MINUTES = 5
MAX_ATTEMPTS = 8
BACKFILL_GAP = 5 # seconds between backfill checks, across workers
SILO_NAMES = {"gotosocial": "GoToSocial", "bluesky": "Bluesky", "threads": "Threads"}
OPT_OUT_RE = re.compile(r"#no(bridge|bot)\b", re.IGNORECASE)
# Silos whose credential was refused this run: target id -> message
denied: Dict[str, str] = {}
@dataclass
class Response:
"""One like, repost, quote or reply on a silo"""
kind: str # like, repost, quote, reply
source_url: str # its URL (likes/reposts: the copy + #liked-by-<id>)
remote_id: str # silo id of the response, or of the actor
author_id: str = ""
author_name: str = ""
author_url: str = ""
author_photo: Optional[str] = None
author_bio: str = ""
opted_out: bool = False # set by the client from silo-specific signals
content_html: Optional[str] = None
published: Optional[str] = None
@dataclass
class Harvest:
"""What one check found. complete=False: some list was cut short."""
responses: List[Response]
complete: bool = True
counts: Optional[Dict[str, int]] = None # Threads likes/reposts/quotes
# --- Text -----------------------------------------------------------------
def text_html(text: str) -> str:
"""Plain text as HTML: escaped, URLs linked, line breaks kept"""
escaped = html.escape(text)
linked = re.sub(r"https?://[^\s<>&\"]+",
lambda m: f'<a href="{m.group(0)}">{m.group(0)}</a>', escaped)
return "<p>" + linked.replace("\n", "<br>") + "</p>" if text.strip() else ""
def facets_html(text: str, facets: list) -> str:
"""Bluesky post text as HTML, with its link facets (byte ranges) as links"""
data = text.encode()
links = []
for facet in facets:
index = facet.get("index") or {}
for feature in facet.get("features") or []:
uri = feature.get("uri") or ""
if feature.get("$type") == "app.bsky.richtext.facet#link" and uri.startswith(("http://", "https://")):
links.append((index.get("byteStart", 0), index.get("byteEnd", 0), uri))
links.sort()
parts, pos = [], 0
for start, end, uri in links:
if start < pos or end > len(data) or start >= end:
continue
parts.append(html.escape(data[pos:start].decode(errors="ignore")))
label = html.escape(data[start:end].decode(errors="ignore"))
parts.append(f'<a href="{html.escape(uri, quote=True)}">{label}</a>')
pos = end
parts.append(html.escape(data[pos:].decode(errors="ignore")))
body = "".join(parts)
return "<p>" + body.replace("\n", "<br>") + "</p>" if text.strip() else ""
# --- Scheduling --------------------------------------------------------------
def interval_for(age_seconds: float, max_age_days: int) -> Optional[int]:
"""Seconds until a note of this age is checked again; None past the cutoff"""
if age_seconds >= max_age_days * DAY:
return None
for younger_than, every in SCHEDULE:
if age_seconds < younger_than:
return every
return LATE_INTERVAL
def _age_seconds(created_at) -> float:
if isinstance(created_at, str):
created_at = datetime.fromisoformat(created_at.replace("Z", "+00:00"))
if created_at.tzinfo is None:
created_at = created_at.replace(tzinfo=timezone.utc)
return (datetime.now(timezone.utc) - created_at).total_seconds()
_ELIGIBLE = """
FROM syndications s JOIN notes n ON n.id = s.note_id
WHERE s.state = 'synced' AND s.enabled = 1 AND s.remote_id IS NOT NULL
AND n.published = 1 AND n.deleted_at IS NULL
AND n.like_of IS NULL AND n.repost_of IS NULL
"""
def _claim(db, row_id: int) -> bool:
cursor = db.execute(
f"""
UPDATE syndications SET backfeed_claimed_until = datetime('now', '+{CLAIM_MINUTES} minutes')
WHERE id = ? AND (backfeed_claimed_until IS NULL OR backfeed_claimed_until < datetime('now'))
""",
(row_id,),
)
db.commit()
return cursor.rowcount == 1
def _backfill_turn(db) -> bool:
"""Take the next backfill slot: at most one check every BACKFILL_GAP seconds"""
cursor = db.execute(
f"""
UPDATE site_state SET value = datetime('now')
WHERE key = 'backfeed_backfill_last' AND value <= datetime('now', '-{BACKFILL_GAP} seconds')
"""
)
if cursor.rowcount == 0:
cursor = db.execute(
"INSERT OR IGNORE INTO site_state (key, value) VALUES ('backfeed_backfill_last', datetime('now'))")
db.commit()
return cursor.rowcount == 1
def _due(db, limit: int) -> list:
"""Routine checks due now; a backfill check only when none are"""
max_age = int(current_app.config.get("BACKFEED_MAX_AGE_DAYS", 90))
lease = "(s.backfeed_claimed_until IS NULL OR s.backfeed_claimed_until < datetime('now'))"
rows = db.execute(
f"""
SELECT s.*, n.created_at AS note_created, n.slug AS note_slug {_ELIGIBLE}
AND {lease}
AND n.created_at >= datetime('now', ?)
AND (s.backfeed_next_at IS NULL OR s.backfeed_next_at <= datetime('now'))
ORDER BY s.backfeed_next_at IS NOT NULL, s.backfeed_next_at LIMIT ?
""",
(f"-{max_age} days", limit),
).fetchall()
if rows:
return rows
backfill = db.execute(
f"""
SELECT s.*, n.created_at AS note_created, n.slug AS note_slug {_ELIGIBLE}
AND {lease} AND s.backfeed_backfill = 1
AND (s.backfeed_next_at IS NULL OR s.backfeed_next_at <= datetime('now'))
ORDER BY s.id LIMIT 1
"""
).fetchall()
if backfill and _backfill_turn(db):
return backfill
return []
def _finish(db, row) -> None:
max_age = int(current_app.config.get("BACKFEED_MAX_AGE_DAYS", 90))
every = interval_for(_age_seconds(row["note_created"]), max_age)
if every is None:
next_at = None
else:
next_at = f"+{int(every * random.uniform(1.0, 1.1))} seconds"
db.execute(
"""
UPDATE syndications
SET backfeed_checked_at = datetime('now'), backfeed_claimed_until = NULL,
backfeed_attempts = 0, backfeed_error = NULL, backfeed_backfill = 0,
backfeed_next_at = CASE WHEN ? IS NULL THEN NULL ELSE datetime('now', ?) END
WHERE id = ?
""",
(next_at, next_at, row["id"]),
)
db.commit()
def _fail(db, row, message: str) -> None:
"""Back off; after MAX_ATTEMPTS drop any backfill and try again in a month"""
attempts = row["backfeed_attempts"] + 1
stop = attempts >= MAX_ATTEMPTS
delay = 30 * DAY if stop else min(30 * (2 ** attempts), 6 * 3600)
db.execute(
f"""
UPDATE syndications
SET backfeed_attempts = ?, backfeed_error = ?, backfeed_claimed_until = NULL,
backfeed_backfill = CASE WHEN ? THEN 0 ELSE backfeed_backfill END,
backfeed_next_at = datetime('now', '+{delay} seconds')
WHERE id = ?
""",
(attempts, message[:500], 1 if stop else 0, row["id"]),
)
db.commit()
current_app.logger.warning(
f"Backfeed {row['target']} for note {row['note_id']} failed (attempt {attempts}): {message}")
# --- Storing -----------------------------------------------------------------
def _opted_out(response: Response) -> bool:
return response.opted_out or bool(OPT_OUT_RE.search(html.unescape(response.author_bio or "")))
def _display_state(db, note_id: int) -> list:
return [tuple(r) for r in db.execute(
"""
SELECT id, mention_type, author_name, author_url, author_photo, content_snippet
FROM webmentions
WHERE target_note_id = ? AND validation_status = 'valid' AND moderation_status = 'approved'
ORDER BY id
""",
(note_id,),
)]
def store(db, note_id: int, target_url: str, origin: str, harvest: Harvest) -> None:
"""
Upsert a check's responses for one note and silo. With a complete
harvest, that silo's responses for the note that weren't found are
marked invalid. Responses from people who opted out are removed.
"""
from starpunk.webmention_processor import _snippet
moderation = "approved" if current_app.config.get("WEBMENTION_AUTO_APPROVE", True) else "pending"
silo = SILO_NAMES.get(origin, origin)
keep = [r for r in harvest.responses if not _opted_out(r)]
dropped = [r.source_url for r in harvest.responses if _opted_out(r)]
from starpunk.avatars import cached_name
before = _display_state(db, note_id)
for r in keep:
snippet = _snippet(r.content_html) if r.content_html else None
# Threads avatar URLs are signed and change on every request: keep
# the stored URL while its saved copy exists, so it isn't refetched
old = db.execute("SELECT author_photo FROM webmentions WHERE source_url = ? AND target_url = ?",
(r.source_url, target_url)).fetchone()
if old and old["author_photo"] and origin == "threads" \
and cached_name(current_app.config["DATA_PATH"], old["author_photo"]):
r.author_photo = old["author_photo"]
db.execute(
"""
INSERT INTO webmentions (
source_url, target_url, target_note_id, mention_type, author_name, author_url,
author_photo, content_snippet, validation_status, moderation_status,
validated_at, source_published, origin, remote_id)
VALUES (?, ?, ?, ?, ?, ?, ?, ?, 'valid', ?, CURRENT_TIMESTAMP, ?, ?, ?)
ON CONFLICT(source_url, target_url) DO UPDATE SET
target_note_id = excluded.target_note_id,
mention_type = excluded.mention_type,
author_name = excluded.author_name,
author_url = excluded.author_url,
author_photo = excluded.author_photo,
content_snippet = excluded.content_snippet,
source_published = COALESCE(excluded.source_published, webmentions.source_published),
origin = excluded.origin,
remote_id = excluded.remote_id,
validation_status = 'valid',
validation_error = NULL,
validated_at = CURRENT_TIMESTAMP,
updated_at = CURRENT_TIMESTAMP
""",
(r.source_url, target_url, note_id, r.kind, r.author_name or None, r.author_url or None,
r.author_photo or None, snippet, moderation, r.published, origin, r.remote_id),
)
if dropped:
marks = ",".join("?" * len(dropped))
db.execute(f"DELETE FROM webmentions WHERE origin = ? AND target_note_id = ? AND source_url IN ({marks})",
[origin, note_id] + dropped)
if harvest.complete:
found = [r.source_url for r in keep]
marks = ",".join("?" * len(found)) or "''"
db.execute(
f"""
UPDATE webmentions
SET validation_status = 'invalid', validation_error = ?, updated_at = CURRENT_TIMESTAMP
WHERE target_note_id = ? AND origin = ? AND validation_status = 'valid'
AND source_url NOT IN ({marks})
""",
[f"No longer on {silo}", note_id, origin] + found,
)
db.commit()
if _display_state(db, note_id) != before:
try:
from pathlib import Path
from starpunk.webmention_sender import note_responses_changed
note_responses_changed(Path(current_app.config["DATABASE_PATH"]), [note_id])
except Exception as e:
current_app.logger.warning(f"Failed to re-send webmentions for note {note_id}: {e}")
from starpunk import avatars
for r in keep:
avatars.fetch(current_app.config["DATA_PATH"], r.author_photo,
current_app.config.get("SITE_URL", ""))
def _store_counts(db, row_id: int, counts: Optional[dict]) -> None:
if counts is None:
return
db.execute(
"UPDATE syndications SET likes_count = ?, reposts_count = ?, quotes_count = ? WHERE id = ?",
(counts.get("likes"), counts.get("reposts"), counts.get("quotes"), row_id),
)
db.commit()
# --- Checking ---------------------------------------------------------------
def _is_denied(error) -> bool:
"""The silo refused the credential (missing scope or bad token)"""
from starpunk.syndication import _GRAPH_AUTH
return getattr(error, "status", None) in (401, 403) or getattr(error, "code", None) in _GRAPH_AUTH
SCOPES = {
"gotosocial": "a GoToSocial token with read:statuses and read:accounts (or read)",
"bluesky": "a working Bluesky app password",
"threads": "a Threads token with threads_read_replies and threads_manage_insights",
}
def check(row, target) -> None:
"""Check one copy now and store what it found (raises SyndicationError)"""
db = get_db(current_app)
site = current_app.config["SITE_URL"].rstrip("/")
harvest = target.client.responses(row["remote_id"], row["remote_url"] or "")
store(db, row["note_id"], f"{site}/note/{row['note_slug']}", row["target"], harvest)
_store_counts(db, row["id"], harvest.counts)
def process_due(limit: int = 3) -> int:
"""Check up to `limit` copies that are due. Returns how many were handled."""
if not current_app.config.get("BACKFEED_ENABLED", False):
return 0
from starpunk.syndication import SyndicationError, get_targets
targets = get_targets()
db = get_db(current_app)
handled = 0
for row in _due(db, limit):
if row["target"] in denied or row["target"] not in targets:
continue
if not _claim(db, row["id"]):
continue
handled += 1
try:
check(row, targets[row["target"]])
except SyndicationError as e:
if _is_denied(e):
denied[row["target"]] = f"Needs {SCOPES.get(row['target'], 'read access')}: {e}"
current_app.logger.error(f"Backfeed for {row['target']} stopped: {denied[row['target']]}")
db.execute("UPDATE syndications SET backfeed_claimed_until = NULL WHERE id = ?", (row["id"],))
db.commit()
continue
_fail(db, row, str(e))
continue
except Exception as e: # one bad copy mustn't stop the worker
_fail(db, row, f"unexpected error: {e}")
continue
_finish(db, row)
_note_backfill_done(db)
return handled
# --- Backfill and admin ---------------------------------------------------------
def _state(db, key: str) -> Optional[str]:
row = db.execute("SELECT value FROM site_state WHERE key = ?", (key,)).fetchone()
return row["value"] if row and row["value"] else None
def _set_state(db, key: str, value: str) -> None:
db.execute("INSERT INTO site_state (key, value) VALUES (?, ?) "
"ON CONFLICT(key) DO UPDATE SET value = excluded.value", (key, value))
def start_backfill() -> int:
"""Mark every eligible copy for one check, whatever its age. Returns how many."""
db = get_db(current_app)
cursor = db.execute(
f"""
UPDATE syndications
SET backfeed_backfill = 1, backfeed_attempts = 0, backfeed_next_at = datetime('now')
WHERE id IN (SELECT s.id {_ELIGIBLE})
"""
)
now = datetime.now(timezone.utc).strftime("%Y-%m-%d %H:%M:%S")
_set_state(db, "backfeed_backfill_started", now)
_set_state(db, "backfeed_backfill_finished", "")
db.commit()
return cursor.rowcount
def _note_backfill_done(db) -> None:
if _state(db, "backfeed_backfill_started") and not _state(db, "backfeed_backfill_finished"):
left = db.execute("SELECT COUNT(*) FROM syndications WHERE backfeed_backfill = 1").fetchone()[0]
if left == 0:
_set_state(db, "backfeed_backfill_finished",
datetime.now(timezone.utc).strftime("%Y-%m-%d %H:%M:%S"))
db.commit()
def check_now(syndication_id: int) -> None:
"""Check this copy's responses at the next worker pass"""
db = get_db(current_app)
db.execute("UPDATE syndications SET backfeed_next_at = datetime('now'), backfeed_attempts = 0 "
"WHERE id = ?", (syndication_id,))
db.commit()
def status() -> dict:
"""For the admin pages: backfill progress and silos whose credential was refused"""
db = get_db(current_app)
return {
"enabled": bool(current_app.config.get("BACKFEED_ENABLED", False)),
"started": _state(db, "backfeed_backfill_started"),
"finished": _state(db, "backfeed_backfill_finished"),
"remaining": db.execute(
"SELECT COUNT(*) FROM syndications WHERE backfeed_backfill = 1").fetchone()[0],
"denied": dict(denied),
"errors": db.execute(
"SELECT COUNT(*) FROM syndications WHERE backfeed_error IS NOT NULL").fetchone()[0],
}
+3
View File
@@ -146,6 +146,9 @@ def load_config(app, config_override=None):
app.config["THREADS_MAX_CHARS"] = int(os.getenv("THREADS_MAX_CHARS", "500"))
app.config["THREADS_MAX_MEDIA"] = int(os.getenv("THREADS_MAX_MEDIA", "4"))
app.config["SYNDICATION_POLL_INTERVAL"] = int(os.getenv("SYNDICATION_POLL_INTERVAL", "10"))
# Backfeed of responses to crossposts (v2.13.0, ADR-073)
app.config["BACKFEED_ENABLED"] = os.getenv("BACKFEED_ENABLED", "false").lower() == "true"
app.config["BACKFEED_MAX_AGE_DAYS"] = int(os.getenv("BACKFEED_MAX_AGE_DAYS", "90"))
# Apply overrides if provided
# CSRF protection for admin and consent forms (v2.11.0); on by default.
+31 -2
View File
@@ -5,7 +5,7 @@ Handles authenticated admin functionality including dashboard, note creation,
editing, and deletion. All routes require authentication.
"""
from flask import Blueprint, flash, g, jsonify, redirect, render_template, request, url_for
from flask import Blueprint, current_app, flash, g, jsonify, redirect, render_template, request, url_for
import logging
import os
import shutil
@@ -259,6 +259,7 @@ def edit_note_form(note_id: int):
syndication_targets=get_targets().values(),
syndications=syndications,
syndication_enabled=enabled,
backfeed_enabled=current_app.config.get("BACKFEED_ENABLED", False),
)
@@ -496,18 +497,24 @@ def webmentions():
from starpunk.webmention_queries import list_webmentions
status_filter = request.args.get('status')
origin_filter = request.args.get('origin') or None
page = int(request.args.get('page', 1))
webmentions, total_pages = list_webmentions(
status_filter=status_filter,
page=page,
per_page=25
per_page=25,
origin_filter=origin_filter,
)
from starpunk import backfeed
return render_template(
"admin/webmentions.html",
webmentions=webmentions,
status_filter=status_filter,
origin_filter=origin_filter,
backfeed=backfeed.status(),
page=page,
total_pages=total_pages,
user_me=g.me
@@ -674,6 +681,28 @@ def revoke_token(token_id):
return redirect(url_for("admin.tokens"))
@bp.route("/backfeed/backfill", methods=["POST"])
@require_auth
def backfeed_backfill():
"""Check every crosspost once for responses, whatever its age (v2.13.0)"""
from starpunk import backfeed
count = backfeed.start_backfill()
flash(f"Backfill started: {count} crosspost(s) to check", "success")
return redirect(url_for("admin.webmentions"))
@bp.route("/syndication/<int:syndication_id>/check", methods=["POST"])
@require_auth
def check_syndication_responses(syndication_id):
"""Check one crosspost's responses at the next worker pass (v2.13.0)"""
from starpunk import backfeed
backfeed.check_now(syndication_id)
flash("Responses will be checked shortly", "success")
return redirect(request.referrer or url_for("admin.dashboard"))
@bp.route("/syndication/<int:syndication_id>/retry", methods=["POST"])
@require_auth
def retry_syndication(syndication_id):
+14
View File
@@ -250,6 +250,20 @@ def media_file(path):
return response
@bp.route("/avatars/<name>")
def avatar(name: str):
"""A saved copy of a response author's avatar (v2.13.0, ADR-073)"""
from pathlib import Path
from starpunk.avatars import NAME_RE, folder
if not NAME_RE.match(name):
abort(404)
response = send_from_directory(folder(Path(current_app.config.get('DATA_PATH', 'data'))), name)
# Named by the hash of the original URL; a changed avatar gets a new name
response.headers['Cache-Control'] = 'public, max-age=604800'
return response
@bp.route("/")
def index():
"""
+3
View File
@@ -258,7 +258,10 @@ def receive_webmention():
ON CONFLICT(source_url, target_url) DO UPDATE SET
validation_status = 'pending',
updated_at = CURRENT_TIMESTAMP
WHERE webmentions.origin = 'webmention'
""", (source, target, target_note_id))
# (A backfed response with the same URLs is StarPunk's own copy of
# what the silo said; a webmention doesn't overwrite it. v2.13.0)
db.commit()
current_app.logger.info(
+211 -5
View File
@@ -67,8 +67,9 @@ class MastodonClient:
def _call(self, method: str, path: str, **kwargs) -> httpx.Response:
headers = {**self.headers, **kwargs.pop("headers", {})}
url = path if path.startswith(("http://", "https://")) else self.base_url + path
try:
response = self.http.request(method, self.base_url + path, headers=headers, **kwargs)
response = self.http.request(method, url, headers=headers, **kwargs)
except httpx.HTTPError as e:
raise SyndicationError(f"{method} {path}: {e}")
if response.status_code == 404:
@@ -76,10 +77,12 @@ class MastodonClient:
if response.status_code >= 400:
# 4xx other than rate limiting won't fix itself
permanent = 400 <= response.status_code < 500 and response.status_code != 429
raise SyndicationError(
exc = SyndicationError(
f"{method} {path}: HTTP {response.status_code} {response.text[:200]}",
permanent=permanent,
)
exc.status = response.status_code
raise exc
return response
def upload_media(self, path: Path, mime_type: str, description: str) -> str:
@@ -166,6 +169,61 @@ class MastodonClient:
text, media_ids = self._prepare(post)
return self.update_status(remote_id, text, media_ids)
# --- Backfeed (v2.13.0, ADR-073) ---
def _account_id(self) -> str:
"""Id of the account StarPunk posts as (its own responses are skipped)"""
if not getattr(self, "account_id", None):
self.account_id = self._call("GET", "/api/v1/accounts/verify_credentials").json()["id"]
return self.account_id
def _list(self, path: str, cap: int) -> tuple:
"""Items of a paginated list (Link: rel="next"). Returns (items, complete)."""
items, url, params = [], path, {"limit": 80}
while url:
response = self._call("GET", url, params=params)
items.extend(response.json() or [])
if len(items) >= cap:
return items[:cap], False
url, params = (response.links.get("next") or {}).get("url"), None
return items, True
@staticmethod
def _author(account: dict) -> dict:
return {
"author_id": account.get("id") or "",
"author_name": account.get("display_name") or account.get("username") or account.get("acct") or "",
"author_url": account.get("url") or "",
"author_photo": account.get("avatar_static") or account.get("avatar"),
"author_bio": account.get("note") or "",
}
def responses(self, remote_id: str, remote_url: str, cap: int = 1000):
"""Likes, boosts and public direct replies of one of our statuses"""
from starpunk.backfeed import Harvest, Response
me = self._account_id()
found, complete = [], True
for kind, path, fragment in (("like", "favourited_by", "liked-by"),
("repost", "reblogged_by", "reposted-by")):
accounts, done = self._list(f"/api/v1/statuses/{remote_id}/{path}", cap)
complete = complete and done
for account in accounts:
if account.get("id") == me:
continue
found.append(Response(kind=kind, source_url=f"{remote_url}#{fragment}-{account['id']}",
remote_id=account["id"], **self._author(account)))
context = self._call("GET", f"/api/v1/statuses/{remote_id}/context").json() or {}
for status in context.get("descendants") or []:
account = status.get("account") or {}
if (status.get("in_reply_to_id") != remote_id or status.get("visibility") != "public"
or account.get("id") == me):
continue
found.append(Response(kind="reply", source_url=status.get("url") or status["uri"],
remote_id=status["id"], content_html=status.get("content") or "",
published=status.get("created_at"), **self._author(account)))
return Harvest(found, complete)
def delete(self, remote_id: str) -> None:
action, _, status_id = remote_id.partition(":")
if action in self.NATIVE and status_id:
@@ -227,9 +285,11 @@ class BlueskyClient:
if response.status_code < 400:
return
permanent = 400 <= response.status_code < 500 and response.status_code != 429
raise SyndicationError(
exc = SyndicationError(
f"{nsid}: HTTP {response.status_code} {response.text[:200]}", permanent=permanent
)
exc.status = response.status_code
raise exc
def _login(self) -> None:
response = None
@@ -405,6 +465,94 @@ class BlueskyClient:
def update(self, remote_id: str, post: "Post") -> dict:
raise NotImplementedError("Bluesky posts can't be edited")
# --- Backfeed (v2.13.0, ADR-073) ---
def _paged(self, nsid: str, params: dict, key: str, cap: int) -> tuple:
"""Items of a cursor-paginated XRPC list. Returns (items, complete)."""
items, cursor = [], None
while True:
query = {**params, "limit": 100}
if cursor:
query["cursor"] = cursor
data = self._call("GET", nsid, params=query).json()
page = data.get(key) or []
items.extend(page)
if len(items) >= cap:
return items[:cap], False
cursor = data.get("cursor")
if not cursor or not page:
return items, True
@staticmethod
def _author(profile: dict) -> dict:
handle = profile.get("handle") or profile.get("did") or ""
return {
"author_id": profile.get("did") or "",
"author_name": profile.get("displayName") or handle,
"author_url": f"https://bsky.app/profile/{handle}",
"author_photo": profile.get("avatar"),
"author_bio": profile.get("description") or "",
# The author asked not to be shown to people who aren't logged in
"opted_out": any(label.get("val") == "!no-unauthenticated"
for label in profile.get("labels") or []),
}
def _post_response(self, kind: str, view: dict):
from starpunk.backfeed import Response, facets_html
author = view.get("author") or {}
record = view.get("record") or {}
rkey = view["uri"].rsplit("/", 1)[1]
return Response(
kind=kind,
source_url=f"https://bsky.app/profile/{author.get('handle') or author.get('did')}/post/{rkey}",
remote_id=view["uri"],
content_html=facets_html(record.get("text") or "", record.get("facets") or []),
published=record.get("createdAt") or view.get("indexedAt"),
**self._author(author),
)
def responses(self, remote_id: str, remote_url: str, cap: int = 1000):
"""Likes, reposts, quotes and direct replies of one of our posts"""
from starpunk.backfeed import Harvest, Response
if not self.did:
with self._lock:
self._login()
found, complete = [], True
likes, done = self._paged("app.bsky.feed.getLikes", {"uri": remote_id}, "likes", cap)
complete = complete and done
for like in likes:
actor = like.get("actor") or {}
found.append(Response(kind="like", source_url=f"{remote_url}#liked-by-{actor.get('did')}",
remote_id=actor.get("did") or "", published=like.get("createdAt"),
**self._author(actor)))
reposters, done = self._paged("app.bsky.feed.getRepostedBy", {"uri": remote_id},
"repostedBy", cap)
complete = complete and done
for actor in reposters:
found.append(Response(kind="repost", source_url=f"{remote_url}#reposted-by-{actor.get('did')}",
remote_id=actor.get("did") or "", **self._author(actor)))
quotes, done = self._paged("app.bsky.feed.getQuotes", {"uri": remote_id}, "posts", cap)
complete = complete and done
posts = [self._post_response("quote", view) for view in quotes]
thread = self._call("GET", "app.bsky.feed.getPostThread",
params={"uri": remote_id, "depth": 1, "parentHeight": 0}).json()
for item in (thread.get("thread") or {}).get("replies") or []:
if item.get("post"): # skips blocked and deleted replies
posts.append(self._post_response("reply", item["post"]))
# Post views carry no bio; fetch them to honour #nobridge
dids = list({p.author_id for p in posts if p.author_id})
bios = {}
for i in range(0, len(dids), 25):
profiles = self._call("GET", "app.bsky.actor.getProfiles",
params={"actors": dids[i:i + 25]}).json().get("profiles") or []
bios.update({p["did"]: p.get("description") or "" for p in profiles})
for post in posts:
post.author_bio = bios.get(post.author_id, "")
found.extend(posts)
return Harvest([r for r in found if r.author_id != self.did], complete)
def delete(self, remote_id: str) -> None:
# at://<did>/app.bsky.feed.post/<rkey>; deleting a missing record succeeds
repo, collection, rkey = remote_id[len("at://"):].split("/", 2)
@@ -513,7 +661,7 @@ def page_card(http: httpx.Client, url: str) -> dict:
return card
THREADS_API = "https://graph.threads.net/v1.0"
THREADS_API = "https://graph.threads.com/v1.0" # Meta's current host (v2.13.0)
THREADS_IMAGE_TYPES = {"image/jpeg", "image/png"}
THREADS_IMAGE_LIMIT = 8_000_000
THREADS_MAX_LINKS = 5
@@ -775,6 +923,57 @@ class ThreadsClient:
def update(self, remote_id: str, post: "Post") -> dict:
raise NotImplementedError("Threads posts can't be edited through the API")
# --- Backfeed (v2.13.0, ADR-073) ---
REPLY_FIELDS = "id,text,username,permalink,timestamp,profile_picture_url,hide_status"
SHOWN = (None, "NOT_HUSHED", "UNHUSHED") # hide_status values of replies nobody hid
def responses(self, remote_id: str, remote_url: str, cap: int = 1000):
"""
Direct replies to one of our posts, and its like/repost/quote counts.
The API doesn't say who liked, reposted or quoted (ADR-073 Decision 7).
"""
from starpunk.backfeed import Harvest, Response, text_html
self._me()
replies, after, complete = [], None, True
while True:
params = {"fields": self.REPLY_FIELDS}
if after:
params["after"] = after
data = self._call("GET", f"{remote_id}/replies", **params)
page = data.get("data") or []
replies.extend(page)
if len(replies) >= cap:
replies, complete = replies[:cap], False
break
paging = data.get("paging") or {}
after = (paging.get("cursors") or {}).get("after")
if not page or not after or not paging.get("next"):
break
found = []
for reply in replies:
username = reply.get("username")
# Private accounts have no username or permalink: nothing public to show
if (not username or not reply.get("permalink") or username == self.username
or reply.get("hide_status") not in self.SHOWN):
continue
found.append(Response(
kind="reply", source_url=reply["permalink"], remote_id=reply["id"],
author_id=username, author_name=username,
author_url=f"https://www.threads.com/@{username}",
author_photo=reply.get("profile_picture_url"),
content_html=text_html(reply.get("text") or ""),
published=reply.get("timestamp"),
))
insights = self._call("GET", f"{remote_id}/insights", metric="likes,reposts,quotes")
counts = {}
for metric in insights.get("data") or []:
values = metric.get("values") or [{}]
counts[metric.get("name")] = int((values[0] or {}).get("value") or 0)
return Harvest(found, complete, counts={k: counts.get(k, 0)
for k in ("likes", "reposts", "quotes")})
def delete(self, remote_id: str) -> None:
try:
self._call("DELETE", remote_id)
@@ -1381,12 +1580,19 @@ class SyndicationWorker(threading.Thread):
def run(self):
while not self._stop_event.is_set():
handled = 0
try:
with self.app.app_context():
handled = process_due()
except Exception as e:
self.app.logger.error(f"Syndication worker error: {e}", exc_info=True)
handled = 0
# Responses to the copies (v2.13.0, ADR-073), after any crossposts
try:
from starpunk import backfeed
with self.app.app_context():
handled += backfeed.process_due()
except Exception as e:
self.app.logger.error(f"Backfeed error: {e}", exc_info=True)
if handled == 0:
self._stop_event.wait(self.poll_interval)
+34 -10
View File
@@ -39,7 +39,8 @@ def get_webmentions_for_note(note_id: int) -> Dict[str, List[Dict[str, Any]]]:
note_id: Note database ID
Returns:
Dict with keys: 'replies', 'likes', 'reposts', 'bookmarks', 'mentions'
Dict with keys: 'replies', 'likes', 'reposts', 'bookmarks', 'quotes'
(backfed quote posts, v2.13.0), 'mentions'
Each value is a list of webmention dicts
"""
db = get_db()
@@ -56,24 +57,34 @@ def get_webmentions_for_note(note_id: int) -> Dict[str, List[Dict[str, Any]]]:
WHEN 'like' THEN 2
WHEN 'repost' THEN 3
WHEN 'bookmark' THEN 4
WHEN 'mention' THEN 5
WHEN 'quote' THEN 5
WHEN 'mention' THEN 6
END,
received_at ASC
"""
rows = db.execute(query, (note_id,)).fetchall()
# Backfed responses say which silo they came from (v2.13.0)
silo_names = {}
if any(row['origin'] != 'webmention' for row in rows):
from starpunk.backfeed import SILO_NAMES
from starpunk.syndication import get_targets
silo_names = {**SILO_NAMES, **{t.id: t.name for t in get_targets().values()}}
# Group by type
grouped = {
'replies': [],
'likes': [],
'reposts': [],
'bookmarks': [],
'quotes': [],
'mentions': []
}
for row in rows:
wm = dict(row)
wm['via'] = silo_names.get(wm['origin']) if wm['origin'] != 'webmention' else None
# Parse timestamps (source_published may be missing or malformed)
for key in ('received_at', 'source_published'):
if isinstance(wm.get(key), str):
@@ -91,6 +102,8 @@ def get_webmentions_for_note(note_id: int) -> Dict[str, List[Dict[str, Any]]]:
grouped['reposts'].append(wm)
elif mention_type == 'bookmark':
grouped['bookmarks'].append(wm)
elif mention_type == 'quote':
grouped['quotes'].append(wm)
else:
grouped['mentions'].append(wm)
@@ -100,15 +113,18 @@ def get_webmentions_for_note(note_id: int) -> Dict[str, List[Dict[str, Any]]]:
def list_webmentions(
status_filter: Optional[str] = None,
page: int = 1,
per_page: int = 25
per_page: int = 25,
origin_filter: Optional[str] = None,
) -> tuple[List[Dict[str, Any]], int]:
"""
List webmentions for admin with pagination and filtering.
Args:
status_filter: Filter by moderation_status ('pending', 'approved', 'hidden')
status_filter: Filter by moderation_status ('pending', 'approved', 'hidden',
'deleted'). Deleted ones are left out unless asked for (v2.13.0).
page: Page number (1-indexed)
per_page: Items per page (default: 25)
origin_filter: 'webmention' or a backfeed silo id (v2.13.0)
Returns:
Tuple of (webmentions list, total_pages)
@@ -124,12 +140,16 @@ def list_webmentions(
count_query = "SELECT COUNT(*) FROM webmentions"
params = []
where_clause = ""
conditions = ["moderation_status = ?" if status_filter else "moderation_status != 'deleted'"]
if status_filter:
where_clause = " WHERE w.moderation_status = ?"
count_query += " WHERE moderation_status = ?"
params.append(status_filter)
if origin_filter:
conditions.append("origin = ?")
params.append(origin_filter)
where = " AND ".join(conditions)
where_clause = " WHERE " + where.replace("moderation_status", "w.moderation_status").replace(
"origin", "w.origin")
count_query += " WHERE " + where
# Count total
total_count = db.execute(count_query, params).fetchone()[0]
@@ -206,7 +226,7 @@ def count_webmentions(status_filter: Optional[str] = None) -> int:
query = "SELECT COUNT(*) FROM webmentions WHERE moderation_status = ?"
result = db.execute(query, (status_filter,)).fetchone()
else:
query = "SELECT COUNT(*) FROM webmentions"
query = "SELECT COUNT(*) FROM webmentions WHERE moderation_status != 'deleted'"
result = db.execute(query).fetchone()
return result[0] if result else 0
@@ -260,6 +280,9 @@ def delete_webmentions(webmention_ids: List[int]) -> int:
"""
Delete webmentions by IDs.
The row stays as a tombstone (moderation_status 'deleted', v2.13.0), so
a re-sent webmention or the next backfeed check doesn't bring it back.
Args:
webmention_ids: List of webmention IDs to delete
@@ -276,7 +299,8 @@ def delete_webmentions(webmention_ids: List[int]) -> int:
)
placeholders = ','.join('?' * len(webmention_ids))
query = f"DELETE FROM webmentions WHERE id IN ({placeholders})"
query = (f"UPDATE webmentions SET moderation_status = 'deleted', updated_at = CURRENT_TIMESTAMP "
f"WHERE id IN ({placeholders}) AND moderation_status != 'deleted'")
cursor = db.execute(query, webmention_ids)
db.commit()
+32 -1
View File
@@ -82,6 +82,8 @@ class WebmentionWorker(threading.Thread):
self._stop_event = threading.Event()
self._last_processed = None
self._lock_file = None
from starpunk.avatars import Upkeep
self._avatar_upkeep = Upkeep()
def run(self):
"""
@@ -113,6 +115,9 @@ class WebmentionWorker(threading.Thread):
# Phase 3: Process outgoing webmentions
processed += self._process_pending_sends()
# Local copies of response avatars (v2.13.0)
self._avatar_upkeep.run(self.db_path, self.app.config['DATA_PATH'],
self.site_url)
# Update last processed time
if processed > 0:
@@ -281,6 +286,22 @@ class WebmentionWorker(threading.Thread):
source_url, result.get('source_post_url'), result.get('source_uid')) if u}
return bool(copies & source)
def _is_backfed(self, note_id: Optional[int], source_url: str, result: Dict[str, Any]) -> bool:
"""True if the source is a response already stored by backfeed for this note"""
if note_id is None:
return False
conn = sqlite3.connect(self.db_path)
try:
backfed = {normalize_url(row[0]) for row in conn.execute(
"SELECT source_url FROM webmentions WHERE target_note_id = ? AND origin != 'webmention'",
(note_id,),
)}
finally:
conn.close()
source = {normalize_url(u) for u in (
source_url, result.get('source_post_url'), result.get('source_uid')) if u}
return bool(backfed & source)
def _display_state(self, webmention_id: int) -> Optional[tuple]:
"""
What the note page shows for this webmention, or None if hidden
@@ -328,6 +349,11 @@ class WebmentionWorker(threading.Thread):
if self._is_own_syndicated_copy(target_note_id, source_url, result):
raise ValidationError("Source is this note's own syndicated copy")
# A response StarPunk already backfed from the silo (e.g. a reply
# also sent by Bridgy) would show twice (v2.13.0, ADR-073)
if self._is_backfed(target_note_id, source_url, result):
raise ValidationError("Already backfed from the silo")
# Update database with successful validation
conn = sqlite3.connect(self.db_path)
cursor = conn.cursor()
@@ -372,6 +398,10 @@ class WebmentionWorker(threading.Thread):
conn.commit()
conn.close()
# Pages show a local copy of the avatar (v2.13.0, ADR-073)
from starpunk import avatars
avatars.fetch(self.app.config['DATA_PATH'], result['author_photo'], self.site_url)
logger.info(
f"Validated webmention: {source_url} -> {target_url} "
f"(type={result['mention_type']})"
@@ -422,7 +452,8 @@ class WebmentionWorker(threading.Thread):
# All other errors are 'error' status
if transient:
status = 'error'
elif any(x in error.lower() for x in ['does not link', 'returned 4', 'own syndicated copy']):
elif any(x in error.lower() for x in ['does not link', 'returned 4', 'own syndicated copy',
'already backfed']):
status = 'invalid'
else:
status = 'error'
+7
View File
@@ -343,6 +343,13 @@ small { display: block; margin-top: var(--spacing-xs); color: var(--color-text-l
color: var(--color-text-light);
}
/* Backfed responses and silo counts (v2.13.0) */
.webmention-via,
.silo-counts {
font-size: 0.875rem;
color: var(--color-text-light);
}
/* Facepile */
.facepile {
display: flex;
+6
View File
@@ -102,6 +102,12 @@
<button type="submit" class="button button-secondary">Retry now</button>
</form>
{% endif %}
{% if backfeed_enabled and s.state == 'synced' and s.remote_id %}
<br><small>Responses {% if s.backfeed_checked_at %}checked {{ s.backfeed_checked_at }} UTC{% else %}not checked yet{% endif %}{% if s.backfeed_error %}: {{ s.backfeed_error }}{% endif %}</small>
<form action="{{ url_for('admin.check_syndication_responses', syndication_id=s.id) }}" method="POST" style="display: inline;"><input type="hidden" name="csrf_token" value="{{ csrf_token() }}">
<button type="submit" class="button button-secondary">Check now</button>
</form>
{% endif %}
</li>
{% endfor %}
</ul>
+4 -1
View File
@@ -16,6 +16,7 @@
<div class="detail-row">
<strong>Source:</strong>
<a href="{{ webmention.source_url }}" target="_blank">{{ webmention.source_url }}</a>
{% if webmention.origin and webmention.origin != 'webmention' %}<small>(backfed from {{ webmention.origin }})</small>{% endif %}
</div>
<div class="detail-row">
@@ -43,7 +44,9 @@
{% if webmention.author_photo %}
<div class="detail-row">
<strong>Author Photo:</strong>
<img src="{{ webmention.author_photo }}" alt="{{ webmention.author_name or 'Author' }}" width="48" height="48" loading="lazy">
{% set photo = webmention.author_photo|local_avatar %}
{% if photo %}<img src="{{ photo }}" alt="{{ webmention.author_name or 'Author' }}" width="48" height="48" loading="lazy">{% endif %}
<a href="{{ webmention.author_photo }}" rel="noopener" target="_blank">{{ webmention.author_photo|truncate(60) }}</a>
</div>
{% endif %}
+31 -2
View File
@@ -12,6 +12,27 @@
</a>
</div>
{# Backfeed (v2.13.0, ADR-073) #}
{% if backfeed and backfeed.enabled %}
<section class="backfeed-status">
<h3>Backfeed</h3>
{% for silo, message in backfeed.denied.items() %}
<p class="flash flash-error">{{ silo }}: {{ message }}</p>
{% endfor %}
{% if backfeed.remaining %}
<p>Backfill started {{ backfeed.started }} UTC: {{ backfeed.remaining }} crosspost(s) left to check.</p>
{% elif backfeed.finished %}
<p>Last backfill: started {{ backfeed.started }}, finished {{ backfeed.finished }} UTC.</p>
{% else %}
<p>No backfill has been run. It checks every crosspost once for responses, however old the note.</p>
{% endif %}
{% if backfeed.errors %}<p>{{ backfeed.errors }} crosspost(s) had an error at their last check (see each note's edit page).</p>{% endif %}
<form method="POST" action="{{ url_for('admin.backfeed_backfill') }}"><input type="hidden" name="csrf_token" value="{{ csrf_token() }}">
<button type="submit" class="button"{% if backfeed.remaining %} disabled{% endif %}>Backfill responses</button>
</form>
</section>
{% endif %}
{# Filter tabs #}
<div class="filter-tabs">
<a href="{{ url_for('admin.webmentions') }}" class="{% if not status_filter %}active{% endif %}">
@@ -26,6 +47,14 @@
<a href="{{ url_for('admin.webmentions', status='hidden') }}" class="{% if status_filter == 'hidden' %}active{% endif %}">
Hidden
</a>
<a href="{{ url_for('admin.webmentions', status='deleted') }}" class="{% if status_filter == 'deleted' %}active{% endif %}">
Deleted
</a>
</div>
<div class="filter-tabs">
{% for value, label in [(None, 'Any source'), ('webmention', 'Webmentions'), ('gotosocial', 'GoToSocial'), ('bluesky', 'Bluesky'), ('threads', 'Threads')] %}
<a href="{{ url_for('admin.webmentions', status=status_filter, origin=value) }}" class="{% if origin_filter == value %}active{% endif %}">{{ label }}</a>
{% endfor %}
</div>
{% if webmentions %}
@@ -93,11 +122,11 @@
{% if total_pages > 1 %}
<nav class="pagination">
{% if page > 1 %}
<a href="?page={{ page - 1 }}{% if status_filter %}&status={{ status_filter }}{% endif %}" class="button">Previous</a>
<a href="?page={{ page - 1 }}{% if status_filter %}&status={{ status_filter }}{% endif %}{% if origin_filter %}&origin={{ origin_filter }}{% endif %}" class="button">Previous</a>
{% endif %}
<span>Page {{ page }} of {{ total_pages }}</span>
{% if page < total_pages %}
<a href="?page={{ page + 1 }}{% if status_filter %}&status={{ status_filter }}{% endif %}" class="button">Next</a>
<a href="?page={{ page + 1 }}{% if status_filter %}&status={{ status_filter }}{% endif %}{% if origin_filter %}&origin={{ origin_filter }}{% endif %}" class="button">Next</a>
{% endif %}
</nav>
{% endif %}
+8
View File
@@ -115,6 +115,13 @@
</div>
{% endif %}
{% if quotes %}
<div class="webmention-group webmention-quotes">
<h3>{{ quotes|length }} {{ 'Quote' if quotes|length == 1 else 'Quotes' }}</h3>
{% for wm in quotes %}{{ responses.mention(wm) }}{% endfor %}
</div>
{% endif %}
{% if mentions %}
<div class="webmention-group webmention-mentions">
<h3>{{ mentions|length }} {{ 'Mention' if mentions|length == 1 else 'Mentions' }}</h3>
@@ -123,5 +130,6 @@
{% endif %}
</section>
{% endif %}
{{ responses.silo_counts(syndications) }}
</article>
{% endblock %}
+23 -6
View File
@@ -12,11 +12,18 @@
{# The response's own URL: the source's u-url if it gave one #}
{% macro response_url(wm) %}{{ wm.source_post_url or wm.source_url }}{% endmacro %}
{# Where a backfed response came from, e.g. "via Bluesky" (v2.13.0) #}
{% macro via(wm) %}{% if wm.via %}<span class="webmention-via">via {{ wm.via }}</span>{% endif %}{% endmacro %}
{# Author h-card: name, URL and photo. u-url is on an inner link: on the
h-card element itself it would belong to the enclosing h-cite #}
h-card element itself it would belong to the enclosing h-cite.
The photo is this site's saved copy (v2.13.0); until there is one, the
initial is shown instead of loading it from the other site. #}
{% macro author_card(wm, class='', photo_size=48, show_name=True) %}
<span class="p-author h-card {{ class }}"><a class="u-url" href="{{ wm.author_url or wm.source_url }}" title="{{ wm.author_name or 'Anonymous' }}">
{%- if wm.author_photo %}<img class="u-photo" src="{{ wm.author_photo }}" alt="" width="{{ photo_size }}" height="{{ photo_size }}" loading="lazy">
{% set photo = wm.author_photo|local_avatar %}
<span class="p-author h-card {{ class }}"><a class="u-url" href="{{ wm.author_url or wm.source_url }}" title="{{ wm.author_name or 'Anonymous' }}{% if wm.via %} via {{ wm.via }}{% endif %}">
{%- if photo %}<img class="u-photo" src="{{ photo }}" alt="" width="{{ photo_size }}" height="{{ photo_size }}" loading="lazy">
{%- elif wm.author_photo %}<data class="u-photo" value="{{ wm.author_photo }}"></data>{% if not show_name %}<span class="face-initial" style="background-color: {{ wm.author_url|url_to_color }}" aria-hidden="true">{{ (wm.author_name or '?')[0]|upper }}</span>{% endif %}
{%- elif not show_name %}<span class="face-initial" style="background-color: {{ wm.author_url|url_to_color }}" aria-hidden="true">{{ (wm.author_name or '?')[0]|upper }}</span>
{%- endif %}
<span class="p-name{{ '' if show_name else ' visually-hidden' }}">{{ wm.author_name or 'Anonymous' }}</span>
@@ -30,7 +37,7 @@
{% if wm.content_snippet %}
<div class="e-content">{{ wm.content_snippet|safe }}</div>
{% endif %}
<a class="u-url webmention-date" href="{{ response_url(wm) }}">{{ response_time(wm) }}</a>
<a class="u-url webmention-date" href="{{ response_url(wm) }}">{{ response_time(wm) }}</a> {{ via(wm) }}
</article>
{%- endmacro %}
@@ -50,13 +57,23 @@
</div>
{%- endmacro %}
{# A post that mentions this note without replying #}
{# A post that mentions this note without replying; also quote posts (v2.13.0) #}
{% macro mention(wm) %}
<div class="webmention webmention-mention u-mention h-cite">
{{ author_card(wm, photo_size=24) }}
{% if wm.content_snippet %}
<p class="p-content">{{ wm.content_snippet|striptags|truncate(200) }}</p>
{% endif %}
<a class="u-url webmention-date" href="{{ response_url(wm) }}">{{ response_time(wm) }}</a>
<a class="u-url webmention-date" href="{{ response_url(wm) }}">{{ response_time(wm) }}</a> {{ via(wm) }}
</div>
{%- endmacro %}
{# Like, repost and quote counts a silo reports without saying who (Threads,
v2.13.0). Plain text: these aren't people, so no u-like / h-cite. #}
{% macro silo_counts(syndications) %}
{% for s in syndications or [] if s.likes_count or s.reposts_count or s.quotes_count %}
{% set parts = [] %}
{% for n, one, many in [(s.likes_count, 'like', 'likes'), (s.reposts_count, 'repost', 'reposts'), (s.quotes_count, 'quote', 'quotes')] if n %}{% set _ = parts.append(n ~ ' ' ~ (one if n == 1 else many)) %}{% endfor %}
<p class="silo-counts">On {{ s.name }}: {{ parts|join(' · ') }}</p>
{% endfor %}
{%- endmacro %}
+564
View File
@@ -0,0 +1,564 @@
"""
Tests for backfeed of responses to crossposts (v2.13.0, ADR-073)
Silo clients are tested against fake APIs (httpx.MockTransport). The
worker, storage and pages are tested with a fake client that returns
whatever a test puts in it.
"""
import io
import json
from urllib.parse import parse_qs
import httpx
import pytest
from PIL import Image
from starpunk import avatars, backfeed, create_app, syndication
from starpunk.backfeed import Harvest, Response, facets_html, interval_for, text_html
from starpunk.database import get_db
from starpunk.notes import create_note
from starpunk.syndication import BlueskyClient, MastodonClient, SyndicationError, ThreadsClient
SITE = "https://blog.example.com"
GTS = "https://social.example"
COPY = f"{GTS}/@phil/statuses/01COPY"
# --- Fakes --------------------------------------------------------------------
class FakeClient:
"""Stands in for a silo client: returns the harvest a test sets"""
config_token = "token"
def __init__(self):
self.harvest = Harvest([])
self.error = None
self.calls = []
def responses(self, remote_id, remote_url, cap=1000):
self.calls.append(remote_id)
if self.error:
raise self.error
return self.harvest
def maintain(self):
pass
def like(n, bio="", **kw):
return Response(kind="like", source_url=f"{COPY}#liked-by-{n}", remote_id=str(n),
author_id=str(n), author_name=f"Liker {n}", author_url=f"{GTS}/@l{n}",
author_bio=bio, **kw)
def reply(n, text="Nice!"):
return Response(kind="reply", source_url=f"{GTS}/@r{n}/statuses/{n}", remote_id=str(n),
author_id=f"r{n}", author_name=f"Replier {n}", author_url=f"{GTS}/@r{n}",
content_html=f"<p>{text}</p>", published="2026-11-10T12:00:00Z")
def resolve(host, *args, **kwargs):
return [(2, 1, 6, "", ("93.184.216.34", 0))]
def png(color="red", size=(200, 120)):
buf = io.BytesIO()
Image.new("RGB", size, color).save(buf, "PNG")
return buf.getvalue()
class FakeWeb:
"""Avatar images by URL (requests are pinned to an IP; Host names the site)"""
def __init__(self):
self.files = {}
def __call__(self, request):
url = f"{request.url.scheme}://{request.headers['host']}{request.url.raw_path.decode()}"
if url in self.files:
return httpx.Response(200, headers={"content-type": "image/png"}, content=self.files[url])
return httpx.Response(404)
@pytest.fixture
def web(monkeypatch):
fake = FakeWeb()
monkeypatch.setattr("starpunk.webmention_ssrf.TRANSPORT", httpx.MockTransport(fake))
monkeypatch.setattr("starpunk.webmention_ssrf.socket.getaddrinfo", resolve)
return fake
@pytest.fixture
def client():
return FakeClient()
@pytest.fixture
def app(tmp_path, client, web):
data = tmp_path / "data"
app = create_app({
"TESTING": True,
"SITE_URL": SITE,
"ADMIN_ME": "https://example.com",
"SESSION_SECRET": "test-secret",
"DATA_PATH": data,
"NOTES_PATH": data / "notes",
"DATABASE_PATH": data / "starpunk.db",
"GOTOSOCIAL_URL": GTS,
"GOTOSOCIAL_ACCESS_TOKEN": "token",
"BACKFEED_ENABLED": True,
})
syndication._client_cache[GTS] = client
backfeed.denied.clear()
yield app
syndication._client_cache.clear()
backfeed.denied.clear()
def crossposted(app, age_days=0, content="Hello"):
"""A published note with a synced GoToSocial copy; returns (note, syndication id)"""
with app.app_context():
note = create_note(content=content, published=True)
db = get_db(app)
if age_days:
db.execute("UPDATE notes SET created_at = datetime('now', ?) WHERE id = ?",
(f"-{age_days} days", note.id))
cursor = db.execute(
"INSERT INTO syndications (note_id, target, enabled, remote_id, remote_url, state) "
"VALUES (?, 'gotosocial', 1, '01COPY', ?, 'synced')", (note.id, COPY))
db.commit()
return note, cursor.lastrowid
def rows(app, note_id):
with app.app_context():
return [dict(r) for r in get_db(app).execute(
"SELECT * FROM webmentions WHERE target_note_id = ? ORDER BY source_url", (note_id,))]
def run(app):
with app.app_context():
return backfeed.process_due()
# --- Schedule and text ------------------------------------------------------------
def test_schedule_slows_with_age_and_stops_at_cutoff():
day = 24 * 3600
assert interval_for(3600, 90) == 15 * 60
assert interval_for(2 * day, 90) == 3600
assert interval_for(10 * day, 90) == day
assert interval_for(60 * day, 90) == 7 * day
assert interval_for(91 * day, 90) is None
def test_text_and_facets_become_safe_html():
assert text_html("hi <b>\nsee https://a.example/x") == \
'<p>hi &lt;b&gt;<br>see <a href="https://a.example/x">https://a.example/x</a></p>'
text = "café: example.com/long…"
start = len("café: ".encode())
facets = [{"index": {"byteStart": start, "byteEnd": len(text.encode())},
"features": [{"$type": "app.bsky.richtext.facet#link", "uri": "https://example.com/long/path"}]}]
assert facets_html(text, facets) == \
'<p>café: <a href="https://example.com/long/path">example.com/long…</a></p>'
assert facets_html("", []) == ""
# --- Worker and storage -------------------------------------------------------------
def test_check_stores_responses_as_webmentions(app, client):
note, sid = crossposted(app)
client.harvest = Harvest([like(1), reply(2, "Great <script>x</script>post")])
assert run(app) == 1
stored = rows(app, note.id)
assert [r["mention_type"] for r in stored] == ["like", "reply"]
assert all(r["origin"] == "gotosocial" and r["validation_status"] == "valid"
and r["moderation_status"] == "approved" for r in stored)
assert stored[1]["content_snippet"] == "<p>Great post</p>"
assert stored[0]["target_url"] == f"{SITE}/note/{note.slug}"
with app.app_context():
s = get_db(app).execute("SELECT * FROM syndications WHERE id = ?", (sid,)).fetchone()
assert s["backfeed_checked_at"] and s["backfeed_next_at"] and s["backfeed_error"] is None
assert run(app) == 0, "not due again for 15 minutes"
def test_repeat_check_updates_rows_and_keeps_moderation(app, client):
note, sid = crossposted(app)
client.harvest = Harvest([like(1), like(2)])
run(app)
with app.app_context():
db = get_db(app)
db.execute("UPDATE webmentions SET moderation_status = 'hidden' WHERE source_url LIKE '%-1'")
db.execute("UPDATE syndications SET backfeed_next_at = datetime('now')")
db.commit()
client.harvest = Harvest([like(1), like(2)])
run(app)
stored = rows(app, note.id)
assert len(stored) == 2
assert [r["moderation_status"] for r in stored] == ["hidden", "approved"]
def test_withdrawn_responses_drop_off_and_come_back(app, client):
note, _ = crossposted(app)
client.harvest = Harvest([like(1), reply(2)])
run(app)
def recheck(harvest):
with app.app_context():
get_db(app).execute("UPDATE syndications SET backfeed_next_at = datetime('now')")
get_db(app).commit()
client.harvest = harvest
run(app)
return {r["source_url"]: (r["validation_status"], r["validation_error"]) for r in rows(app, note.id)}
# An incomplete list (cap reached, say) removes nothing
assert all(v == ("valid", None) for v in recheck(Harvest([like(1)], complete=False)).values())
state = recheck(Harvest([like(1)]))
assert state[reply(2).source_url] == ("invalid", "No longer on GoToSocial")
assert recheck(Harvest([like(1), reply(2)]))[reply(2).source_url] == ("valid", None)
def test_people_who_opt_out_are_skipped_and_removed(app, client):
note, _ = crossposted(app)
client.harvest = Harvest([like(1), like(2)])
run(app)
with app.app_context():
get_db(app).execute("UPDATE syndications SET backfeed_next_at = datetime('now')")
get_db(app).commit()
client.harvest = Harvest([like(1, bio="<p>Please #NoBridge thanks</p>"), like(2),
like(3, opted_out=True)])
run(app)
assert [r["source_url"] for r in rows(app, note.id)] == [like(2).source_url]
def test_admin_delete_is_a_tombstone_the_next_check_keeps(app, client):
from starpunk.webmention_queries import count_webmentions, delete_webmentions, list_webmentions
note, _ = crossposted(app)
client.harvest = Harvest([like(1)])
run(app)
with app.test_request_context():
wm_id = rows(app, note.id)[0]["id"]
assert delete_webmentions([wm_id]) == 1
get_db(app).execute("UPDATE syndications SET backfeed_next_at = datetime('now')")
get_db(app).commit()
run(app)
assert rows(app, note.id)[0]["moderation_status"] == "deleted"
with app.test_request_context():
assert list_webmentions()[0] == [] and count_webmentions() == 0
assert len(list_webmentions(status_filter="deleted")[0]) == 1
def test_old_notes_wait_for_a_backfill(app, client):
old, sid = crossposted(app, age_days=200)
client.harvest = Harvest([reply(1)])
assert run(app) == 0
with app.app_context():
assert backfeed.start_backfill() == 1
assert backfeed.status()["remaining"] == 1
assert run(app) == 1
assert len(rows(app, old.id)) == 1
with app.app_context():
status = backfeed.status()
s = get_db(app).execute("SELECT * FROM syndications WHERE id = ?", (sid,)).fetchone()
assert status["remaining"] == 0 and status["finished"]
assert s["backfeed_backfill"] == 0 and s["backfeed_next_at"] is None, "past the cutoff: no more checks"
def test_backfill_is_paced(app, client):
crossposted(app, age_days=200)
crossposted(app, age_days=300)
with app.app_context():
backfeed.start_backfill()
assert run(app) == 1
assert run(app) == 0, "the next backfill check waits a few seconds"
with app.app_context():
get_db(app).execute("UPDATE site_state SET value = datetime('now', '-10 seconds') "
"WHERE key = 'backfeed_backfill_last'")
get_db(app).commit()
assert run(app) == 1
def test_routine_checks_skip_likes_reposts_and_unpublished(app, client):
with app.app_context():
liked = create_note(content="", published=True, like_of="https://other.example/post")
draft = create_note(content="Draft", published=False)
db = get_db(app)
for note in (liked, draft):
db.execute("INSERT INTO syndications (note_id, target, enabled, remote_id, remote_url, state) "
"VALUES (?, 'gotosocial', 1, 'x', 'https://s.example/x', 'synced')", (note.id,))
db.commit()
assert run(app) == 0 and client.calls == []
def test_refused_credential_stops_that_silo(app, client):
crossposted(app)
error = SyndicationError("GET /api/v1/statuses/01COPY/favourited_by: HTTP 403 forbidden", permanent=True)
error.status = 403
client.error = error
run(app)
assert "gotosocial" in backfeed.denied and "read:statuses" in backfeed.denied["gotosocial"]
client.error = None
assert run(app) == 0, "not retried until restart"
def test_failures_back_off(app, client):
_, sid = crossposted(app)
client.error = SyndicationError("HTTP 502 bad gateway")
run(app)
with app.app_context():
s = get_db(app).execute("SELECT * FROM syndications WHERE id = ?", (sid,)).fetchone()
assert s["backfeed_attempts"] == 1 and "502" in s["backfeed_error"]
assert run(app) == 0
def test_disabled_does_nothing(app, client):
crossposted(app)
app.config["BACKFEED_ENABLED"] = False
assert run(app) == 0 and client.calls == []
# --- Pages ------------------------------------------------------------------------------
def test_note_page_shows_backfed_responses_quotes_and_counts(app, client, web):
note, sid = crossposted(app)
web.files["https://cdn.example/a.png"] = png()
quote = Response(kind="quote", source_url="https://bsky.app/profile/q.example/post/3k",
remote_id="at://did:plc:q/app.bsky.feed.post/3k", author_name="Quoter",
author_url="https://bsky.app/profile/q.example", content_html="<p>Look at this</p>")
client.harvest = Harvest([like(1, author_photo="https://cdn.example/a.png"), reply(2), quote])
run(app)
with app.app_context():
get_db(app).execute("UPDATE syndications SET likes_count = 12, reposts_count = 1 WHERE id = ?", (sid,))
get_db(app).commit()
page = app.test_client().get(f"/note/{note.slug}").get_data(as_text=True)
assert "via GoToSocial" in page
assert "1 Quote" in page and "Look at this" in page
assert 'class="silo-counts">On GoToSocial: 12 likes · 1 repost<' in page
assert 'src="/avatars/' in page and 'src="https://cdn.example/a.png"' not in page
def test_avatars_are_saved_served_and_cleaned_up(app, web):
url = "https://cdn.example/face.png"
web.files[url] = png(size=(300, 200))
data = app.config["DATA_PATH"]
assert avatars.cached_name(data, url) is None
with app.app_context():
assert avatars.fetch(data, url, SITE)
assert not avatars.fetch(data, "https://cdn.example/missing.png", SITE)
name = avatars.cached_name(data, url)
with Image.open(avatars.folder(data) / name) as image:
assert image.size == (96, 96) and image.format == "WEBP"
response = app.test_client().get(f"/avatars/{name}")
assert response.status_code == 200 and response.mimetype == "image/webp"
assert app.test_client().get("/avatars/../starpunk.db").status_code == 404
# Nothing uses it: the weekly cleanup removes it
assert avatars.remove_unused(app.config["DATABASE_PATH"], data) == 1
def test_webmention_cannot_overwrite_a_backfed_response(app, client):
note, _ = crossposted(app)
client.harvest = Harvest([reply(7)])
run(app)
response = app.test_client().post("/webmention", data={
"source": reply(7).source_url, "target": f"{SITE}/note/{note.slug}"})
assert response.status_code == 202
assert rows(app, note.id)[0]["validation_status"] == "valid"
# --- Silo clients ---------------------------------------------------------------------------
def mastodon_api(request):
path, query = request.url.path, parse_qs(request.url.query.decode())
account = lambda n, **kw: {"id": n, "display_name": f"User {n}", "url": f"{GTS}/@u{n}",
"avatar": f"{GTS}/a/{n}.png", "note": "", **kw}
if path == "/api/v1/accounts/verify_credentials":
return httpx.Response(200, json={"id": "me"})
if path == "/api/v1/statuses/01COPY/favourited_by":
if "max_id" in query:
return httpx.Response(200, json=[account("3")])
return httpx.Response(200, json=[account("1"), account("me")],
headers={"Link": f'<{GTS}/api/v1/statuses/01COPY/favourited_by?max_id=9>; rel="next"'})
if path == "/api/v1/statuses/01COPY/reblogged_by":
return httpx.Response(200, json=[account("2")])
if path == "/api/v1/statuses/01COPY/context":
status = lambda n, parent="01COPY", vis="public", who=None: {
"id": n, "url": f"{GTS}/@u/statuses/{n}", "uri": f"{GTS}/u/{n}", "in_reply_to_id": parent,
"visibility": vis, "content": f"<p>reply {n}</p>", "created_at": "2026-11-10T00:00:00Z",
"account": who or account("r" + n)}
return httpx.Response(200, json={"ancestors": [], "descendants": [
status("10"), status("11", vis="unlisted"), status("12", parent="10"),
status("13", who=account("me"))]})
return httpx.Response(404)
def test_mastodon_client_reads_likes_boosts_and_public_direct_replies():
gts = MastodonClient(GTS, "t", http=httpx.Client(transport=httpx.MockTransport(mastodon_api)))
harvest = gts.responses("01COPY", COPY)
assert harvest.complete
got = [(r.kind, r.source_url) for r in harvest.responses]
assert got == [("like", f"{COPY}#liked-by-1"), ("like", f"{COPY}#liked-by-3"),
("repost", f"{COPY}#reposted-by-2"), ("reply", f"{GTS}/@u/statuses/10")]
assert harvest.responses[-1].content_html == "<p>reply 10</p>"
assert gts.responses("01COPY", COPY, cap=1).complete is False
def test_bluesky_client_reads_likes_reposts_quotes_and_replies():
me, post = "did:plc:me", "at://did:plc:me/app.bsky.feed.post/abc"
def api(request):
nsid = request.url.path.removeprefix("/xrpc/")
query = parse_qs(request.url.query.decode())
profile = lambda d, **kw: {"did": d, "handle": f"{d[8:]}.example", "displayName": d[8:].title(), **kw}
view = lambda d, rkey, text: {"uri": f"at://{d}/app.bsky.feed.post/{rkey}", "author": profile(d),
"record": {"text": text, "createdAt": "2026-11-10T00:00:00Z"}}
if nsid == "com.atproto.server.createSession":
return httpx.Response(200, json={"did": me, "handle": "me.example", "accessJwt": "a", "refreshJwt": "r"})
if nsid == "app.bsky.feed.getLikes":
if query.get("cursor"):
return httpx.Response(200, json={"likes": [{"actor": profile("did:plc:carol")}]})
return httpx.Response(200, json={"cursor": "c1", "likes": [
{"actor": profile("did:plc:alice"), "createdAt": "2026-11-10T01:00:00Z"},
{"actor": profile("did:plc:shy", labels=[{"val": "!no-unauthenticated"}])},
{"actor": profile(me)}]})
if nsid == "app.bsky.feed.getRepostedBy":
return httpx.Response(200, json={"repostedBy": [profile("did:plc:bob")]})
if nsid == "app.bsky.feed.getQuotes":
return httpx.Response(200, json={"posts": [view("did:plc:dan", "q1", "Quoting this")]})
if nsid == "app.bsky.feed.getPostThread":
return httpx.Response(200, json={"thread": {"post": {}, "replies": [
{"post": view("did:plc:eve", "r1", "A reply")},
{"$type": "app.bsky.feed.defs#blockedPost", "uri": "at://x"}]}})
if nsid == "app.bsky.actor.getProfiles":
return httpx.Response(200, json={"profiles": [
profile("did:plc:dan", description="no thanks #nobot"), profile("did:plc:eve")]})
return httpx.Response(400, json={"error": "Unknown"})
bsky = BlueskyClient("https://bsky.example", "me.example", "pw",
http=httpx.Client(transport=httpx.MockTransport(api)))
harvest = bsky.responses(post, "https://bsky.app/profile/me.example/post/abc")
by = {(r.kind, r.author_id): r for r in harvest.responses}
assert set(by) == {("like", "did:plc:alice"), ("like", "did:plc:shy"), ("like", "did:plc:carol"),
("repost", "did:plc:bob"), ("quote", "did:plc:dan"), ("reply", "did:plc:eve")}
assert by[("like", "did:plc:shy")].opted_out
assert by[("quote", "did:plc:dan")].author_bio == "no thanks #nobot"
assert by[("reply", "did:plc:eve")].source_url == "https://bsky.app/profile/eve.example/post/r1"
assert by[("like", "did:plc:alice")].source_url.endswith("#liked-by-did:plc:alice")
assert harvest.complete
def test_threads_client_reads_replies_and_counts():
def api(request):
path = request.url.path
params = parse_qs(request.url.query.decode())
if path.endswith("/me"):
return httpx.Response(200, json={"id": "42", "username": "me"})
if path.endswith("/111/replies"):
assert "profile_picture_url" in params["fields"][0]
return httpx.Response(200, json={"data": [
{"id": "1", "text": "Hi", "username": "ann", "permalink": "https://www.threads.com/@ann/post/A",
"timestamp": "2026-11-10T00:00:00+0000", "hide_status": "NOT_HUSHED"},
{"id": "2", "text": "private"},
{"id": "3", "text": "hidden", "username": "x", "permalink": "p", "hide_status": "HIDDEN"},
{"id": "4", "text": "mine", "username": "me", "permalink": "p"}]})
if path.endswith("/111/insights"):
assert params["metric"] == ["likes,reposts,quotes"]
return httpx.Response(200, json={"data": [
{"name": "likes", "values": [{"value": 12}]}, {"name": "reposts", "values": [{"value": 3}]},
{"name": "quotes", "values": [{"value": 0}]}]})
return httpx.Response(404, json={"error": {"code": 100}})
threads = ThreadsClient("tok", http=httpx.Client(transport=httpx.MockTransport(api)))
harvest = threads.responses("111", "https://www.threads.com/@me/post/X")
assert [(r.kind, r.author_name, r.content_html) for r in harvest.responses] == [("reply", "ann", "<p>Hi</p>")]
assert harvest.counts == {"likes": 12, "reposts": 3, "quotes": 0}
# --- Migration ---------------------------------------------------------------------------
def test_migration_020_keeps_existing_webmentions(tmp_path):
import sqlite3
from pathlib import Path
from starpunk.database.schema import INITIAL_SCHEMA_SQL
from starpunk.migrations import execute_in_transaction
migrations = Path(__file__).parent.parent / "migrations"
conn = sqlite3.connect(tmp_path / "old.db")
conn.execute("PRAGMA foreign_keys=ON")
conn.executescript(INITIAL_SCHEMA_SQL)
for name in ("010_add_webmentions.sql", "012_add_syndications.sql", "014_webmention_source_details.sql"):
matches = list(migrations.glob(name.split("_")[0] + "_*.sql"))
conn.execute("BEGIN")
execute_in_transaction(conn, matches[0].read_text())
conn.commit()
conn.execute("INSERT INTO notes (slug, file_path, published) VALUES ('n', 'n.md', 1)")
conn.execute("INSERT INTO webmentions (source_url, target_url, target_note_id, mention_type, "
"validation_status, moderation_status, source_published) "
"VALUES ('https://a.example/r', 'https://b.example/note/n', 1, 'reply', 'valid', 'hidden', "
"'2026-01-01')")
conn.commit()
conn.execute("BEGIN")
execute_in_transaction(conn, (migrations / "020_backfeed.sql").read_text())
conn.commit()
row = conn.execute("SELECT mention_type, moderation_status, source_published, origin, target_note_id "
"FROM webmentions").fetchone()
assert row == ("reply", "hidden", "2026-01-01", "webmention", 1)
conn.execute("INSERT INTO webmentions (source_url, target_url, mention_type, origin) "
"VALUES ('https://q.example', 'https://b.example/note/n', 'quote', 'bluesky')")
columns = {r[1] for r in conn.execute("PRAGMA table_info(syndications)")}
assert {"backfeed_next_at", "backfeed_backfill", "likes_count"} <= columns
assert conn.execute("PRAGMA foreign_key_check").fetchall() == []
# --- Admin ----------------------------------------------------------------------------------
def test_admin_backfill_and_check_now(app, client):
from starpunk.auth import create_session
app.config.update(ADMIN_ME="https://example.com", CSRF_ENABLED=False)
note, sid = crossposted(app, age_days=200)
with app.test_request_context():
token = create_session("https://example.com")
browser = app.test_client()
browser.set_cookie("starpunk_session", token)
page = browser.get("/admin/webmentions").get_data(as_text=True)
assert "No backfill has been run" in page and "Backfill responses" in page
response = browser.post("/admin/backfeed/backfill", follow_redirects=True)
assert "Backfill started: 1 crosspost(s) to check" in response.get_data(as_text=True)
assert "1 crosspost(s) left to check" in browser.get("/admin/webmentions").get_data(as_text=True)
edit = browser.get(f"/admin/edit/{note.id}").get_data(as_text=True)
assert "Responses not checked yet" in edit and "Check now" in edit
browser.post(f"/admin/syndication/{sid}/check")
with app.app_context():
s = get_db(app).execute("SELECT backfeed_next_at FROM syndications WHERE id = ?", (sid,)).fetchone()
assert s["backfeed_next_at"]
backfeed.denied["gotosocial"] = "Needs a GoToSocial token with read:statuses"
assert "Needs a GoToSocial token" in browser.get("/admin/webmentions?origin=gotosocial").get_data(as_text=True)
def test_threads_avatar_kept_while_its_copy_exists(app, client, web):
note, _ = crossposted(app)
with app.app_context():
get_db(app).execute("UPDATE syndications SET target = 'threads'")
get_db(app).commit()
syndication._client_cache["threads"] = client
app.config["THREADS_ACCESS_TOKEN"] = "token"
web.files["https://cdn.example/p.jpg?sig=1"] = png()
first = reply(1)
first.author_photo = "https://cdn.example/p.jpg?sig=1"
client.harvest = Harvest([first])
run(app)
with app.app_context():
get_db(app).execute("UPDATE syndications SET backfeed_next_at = datetime('now')")
get_db(app).commit()
again = reply(1)
again.author_photo = "https://cdn.example/p.jpg?sig=2" # not fetchable: would be a refetch
client.harvest = Harvest([again])
run(app)
assert rows(app, note.id)[0]["author_photo"] == "https://cdn.example/p.jpg?sig=1"