Mac: fix Hermes approval guardian on Haiku 5.5; Bot Chat folder
Hermes's OpenRouter code treats Claude models it doesn't list as
reasoning-mandatory and drops the guardian's "reasoning off" for Haiku
5.5, which then sometimes spends the 16-token budget thinking and
returns nothing, escalating the command to the user. Send
extra_body.reasoning {effort: none} for the approval task; replays of
the failing commands went 21/21 correct.
Also create ~/hermes, the folder the Hermex Bot Chat session works in.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
1 parent
cde36c8710
commit
b2e045c9f1
4 files changed
+21
-2
No files matched your search
@@ -66,10 +66,14 @@ hermes_settings:
|
||||
# Short-answer tasks turn reasoning off and allow only a few tokens (the
|
||||
# approval guardian: 16, titles: 64). GLM-5.3 Flash can't turn reasoning off
|
||||
# (OpenRouter: reasoning mandatory), spends the budget thinking and returns
|
||||
# nothing, so the guardian escalated every command to you. Haiku 5.5 has
|
||||
# reasoning off by default.
|
||||
# nothing, so the guardian escalated every command to you. Haiku 5.5 can
|
||||
# turn reasoning off, but Hermes's OpenRouter code treats Claude models it
|
||||
# doesn't list as reasoning-mandatory and drops the "off" (hermes-agent
|
||||
# #134761). Haiku then sometimes thinks through the 16 tokens on longer
|
||||
# commands. extra_body.reasoning is sent to OpenRouter as written.
|
||||
auxiliary.approval.provider: openrouter
|
||||
auxiliary.approval.model: anthropic/claude-haiku-5.5
|
||||
auxiliary.approval.extra_body: {reasoning: {effort: none}}
|
||||
auxiliary.title_generation.provider: openrouter
|
||||
auxiliary.title_generation.model: anthropic/claude-haiku-5.5
|
||||
auxiliary.memory_query_rewrite.provider: openrouter
|
||||
|
||||
@@ -30,8 +30,13 @@ Manages cogitator-02, an Apple Silicon Mac on the tailnet (`mac.yml`,
|
||||
memory, disabled toolsets, skill-write approval, deny rules, streaming,
|
||||
voice-note transcription, and models: chat on GPT-6.1 Sol, short side tasks
|
||||
(approval, titles) on Claude Haiku 5.5, longer ones and cron on GLM-5.3 Flash, subagents on DeepSeek V4.1 Flash, `/model` aliases.
|
||||
The approval guardian also gets `extra_body.reasoning: {effort: none}`,
|
||||
because Hermes drops its own "reasoning off" for Haiku 5.5 on OpenRouter.
|
||||
Keys in `hermes_settings_force` are set with `--force` (Hermes reads
|
||||
`agent.reasoning_effort` but `config set` refuses it, hermes-agent #140221).
|
||||
Also creates `~/hermes`, the Bot Chat's folder (the Hermex app's default
|
||||
bot). The session's saved folder was set once by hand; Hermes has no
|
||||
command for it.
|
||||
- **MCP servers and secrets** (also tag `hermes-settings`): MCP servers are
|
||||
`mcp_servers.<name>` entries in `hermes_settings`, each running in a
|
||||
throwaway Podman container. Secrets they need are in `hermes_env_secrets`
|
||||
|
||||
@@ -1,3 +1,12 @@
|
||||
# The Hermex app's Bot Chat works here: its session's saved folder is
|
||||
# mounted into the sandbox at /workspace. Saved once with Hermes's
|
||||
# update_session_cwd, since Hermes has no command for it.
|
||||
- name: Bot Chat workspace folder
|
||||
ansible.builtin.file:
|
||||
path: "{{ hermes_bot_workspace }}"
|
||||
state: directory
|
||||
mode: "0755"
|
||||
|
||||
# Secrets go in first, so servers and settings that reference them
|
||||
# (${VAR}) find them when the gateway restarts.
|
||||
- name: Secrets in Hermes's .env
|
||||
|
||||
@@ -4,3 +4,4 @@ hermes_dashboard_label: local.hermes-dashboard
|
||||
tailscale_cli: /Applications/Tailscale.app/Contents/MacOS/Tailscale
|
||||
hermes_cli: "{{ ansible_user_dir }}/.local/bin/hermes"
|
||||
podman_machine_label: local.podman-machine
|
||||
hermes_bot_workspace: "{{ ansible_user_dir }}/hermes"
|
||||
Reference in new issue
Block a user