Final review: startup probe, error messages, refusal exit, docs

A startup probe that failed with anything but OSError (llama-server not in
router mode, an empty model list) crashed the daemon at boot, against its
own comment; it is caught like the per-request failures now. llama-server's
reasons reach the client: an HTTP error carries its body and an error chunk
mid-stream its message, where the client used to see 'choices'. A daemon
that refuses to start, or fails before serving, exits 1 instead of 0.
assistant.sh no longer mistakes a --test run for the daemon.

AGENTS.md said every section from Hardware to Kokoro was voice-only, which
included findings the daemon still applies; it now names the voice ones.
It also claimed a stopped reply emits nothing after done, while a late ctx
can. The spec lists every error message, the example config no longer
shows a non-default temperature as the default, and the updates tool stops
telling the model not to read lists aloud. The spec is marked implemented.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

1 files changed, 11 insertions, 4 deletions

diff --git a/AGENTS.md b/AGENTS.md
index 45d6e31..e25ec1e 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -10,9 +10,14 @@ canvas (`desktop/assistant/` in the quickshell repo, SUPER+Home). Spec:
 `docs/superpowers/specs/2026-10-16-text-daemon-design.md`.
 
 The voice version is the tag `voice-v1`: push-to-talk on the mic's mute
-button, whisper large-v3-turbo, Kokoro. The sections from "Hardware"
-through "Kokoro-82M" describe that version and its measurements. They are
-kept because a return to voice starts from them.
+button, whisper large-v3-turbo, Kokoro. These sections describe only that
+version, and are kept because a return to voice starts from them:
+"Hardware", "whisper.cpp", and under Findings "Models", "STT backends",
+"Live mic with whisper-stream", both "Push-to-talk" sections,
+"large-v3-turbo", "The loop", "Overlay", "Barge-in", "Wake word trial",
+"TTS" and "Kokoro-82M". The rest still describes the daemon: the tool,
+context-size and tool-call-history findings are rules `assistant.py`
+applies today.
 
 Tools (2026-09-26): read-only shell-command TOML files in
 `~/.config/desktop-assistant/tools/`; `config.toml` there overrides the
@@ -35,7 +40,9 @@ socket (ask) ──> llama-server (LLM, tools) ──> socket (think, token, too
 - **Concurrency:** one reader thread per client, one reply thread per
   question. One lock guards the conversation, the turn counter and the
   database. `say()` checks the turn under it before each emit, and stop
-  bumps the turn under it, so a stopped reply cannot emit after its `done`.
+  bumps the turn under it, so a stopped reply cannot send a token or tool
+  event after its `done`. A late `ctx` can still arrive: `ask()` emits it
+  outside the lock, and its numbers are true regardless.
 - **One daemon:** `serve()` probe-connects first and refuses to start over a socket that answers; `main()` removes the socket only if it served it.
 - **Commands** are type-checked before they run (ids must be `int`: SQLite would match `"1"` to row 1 while the Python comparison with the current id would not) and run under a try, so a failing command answers `error` instead of killing the client's reader.
 - **Client sockets** have a 2 s timeout, not non-blocking mode: a large