Commit Graph

33027 Commits

Author SHA1 Message Date
Peter Steinberger
c910ddac38 test: add Kimi and Qianfan extension coverage 2026-04-21 02:41:26 +01:00
Peter Steinberger
82b8a4aab6 docs(openai): clarify GPT-5 prompt defaults 2026-04-21 02:36:16 +01:00
Peter Steinberger
ab03d4e037 fix(openai): default GPT-5 prompt overlay 2026-04-21 02:36:16 +01:00
Andrii Furmanets
b6a8759b29 fix(web-search): restore SecretRef runtime compatibility for bundled providers (#68424)
Adds missing compatibility runtime path metadata for bundled SecretRef-capable web-search providers and keeps the manifest registry covered by a regression test.\n\nThanks @afurm!
2026-04-21 02:34:24 +01:00
Peter Steinberger
f04185cc70 test: stabilize live media and gateway probes 2026-04-21 02:10:19 +01:00
Peter Steinberger
5ab26a8774 ci: extend checkout fetch timeout 2026-04-21 02:05:26 +01:00
aniaan
c8e5150fd4 feat(moonshot): default to Kimi K2.6 with K2.6-only thinking.keep support (#68816)
Merged via squash.

Prepared head SHA: ed54e02842
Co-authored-by: aniaan <40813941+aniaan@users.noreply.github.com>
Co-authored-by: odysseus0 <8635094+odysseus0@users.noreply.github.com>
Reviewed-by: @odysseus0
2026-04-20 18:04:49 -07:00
Peter Steinberger
a112903802 test: use synthetic auto reply fixtures 2026-04-21 01:49:30 +01:00
poisk
32e8bca02c fix(telegram): honor removeAckAfterReply for status reactions (#68067)
Thanks @poiskgit.
2026-04-21 01:47:20 +01:00
Peter Steinberger
60a1f01a3e test: use synthetic agent infra fixtures 2026-04-21 01:46:33 +01:00
Peter Steinberger
442da01db4 test: use synthetic program status fixtures 2026-04-21 01:44:03 +01:00
Peter Steinberger
969ca8511d test: use synthetic cli provider fixtures 2026-04-21 01:42:29 +01:00
Peter Steinberger
66665eea6d test: use synthetic status session fixtures 2026-04-21 01:40:29 +01:00
Peter Steinberger
6bb6cfc68e test: use synthetic plugin channel fixtures 2026-04-21 01:32:27 +01:00
Peter Steinberger
97e528ed54 test: use synthetic agent session fixtures 2026-04-21 01:24:34 +01:00
Peter Steinberger
9f2f89320e test: use synthetic infra channel fixtures 2026-04-21 01:21:05 +01:00
Peter Steinberger
14ceec27fa test: use synthetic config cron channel fixtures 2026-04-21 01:19:35 +01:00
Peter Steinberger
f50202ee95 test: use synthetic auto-reply channel fixtures 2026-04-21 01:18:05 +01:00
Peter Steinberger
f3b56165f5 docs(telegram): clarify polling stall tuning 2026-04-21 01:15:28 +01:00
Peter Steinberger
e8898bb6c1 test: use synthetic agent channel fixtures 2026-04-21 01:15:11 +01:00
Peter Steinberger
3f274006cd refactor: share oauth callback flow 2026-04-21 01:07:09 +01:00
Peter Steinberger
f85c0b7dc5 refactor: reuse shared local file access 2026-04-21 01:07:09 +01:00
Peter Steinberger
7b1f7b179f refactor: share thread binding lifecycle 2026-04-21 01:07:09 +01:00
Peter Steinberger
4ea8063203 refactor: reuse operator approval gateway lifecycle 2026-04-21 01:07:09 +01:00
Peter Steinberger
6c67339798 docs: note Codex approval default fix (#68721) (thanks @Lucenx9) 2026-04-21 01:06:36 +01:00
Lucenx9
758e83015b docs(codex): clarify approval override example 2026-04-21 01:06:36 +01:00
Lucenx9
d04f7e7ce7 fix(codex): default app-server approvals to on-request 2026-04-21 01:06:36 +01:00
Amine Harch el korane
8c05043eca fix(telegram): tune polling stall threshold
Raise the Telegram polling watchdog default from 90s to 120s and add bounded channels.telegram.pollingStallThresholdMs overrides, including per-account config.\n\nThanks @Vitalcheffe.
2026-04-21 01:03:04 +01:00
Peter Steinberger
660e4257a7 refactor: share codex auth bridge 2026-04-21 00:54:08 +01:00
Peter Steinberger
0647481c7c refactor: share ssrf policy merging 2026-04-21 00:54:08 +01:00
Peter Steinberger
7e28caa637 refactor: share fast mode normalization 2026-04-21 00:54:08 +01:00
Peter Steinberger
44ca47b2eb refactor: share allow-from store file reads 2026-04-21 00:54:08 +01:00
Peter Steinberger
bcd232467f ci: remove channel contract heartbeat 2026-04-21 00:53:50 +01:00
Peter Steinberger
3caf9faef5 test: use synthetic metadata channel labels 2026-04-21 00:52:52 +01:00
Peter Steinberger
71154bf3bf test: use synthetic approval pairing fixtures 2026-04-21 00:51:37 +01:00
Peter Steinberger
05835dd2d4 test: use synthetic heartbeat wake fixtures 2026-04-21 00:50:32 +01:00
Peter Steinberger
25428c4631 fix: keep Codex projector events isolated (#69072) (thanks @ayeshakhalid192007-dev) 2026-04-21 00:49:54 +01:00
ayeshakhalid192007-dev
f2f27775fb fix(codex/app-server): release session lane when projector throws on turn/completed 2026-04-21 00:49:54 +01:00
ayeshakhalid192007-dev
54a2a20447 test(codex): wait for initialize write before reading harness in models.test.ts 2026-04-21 00:49:54 +01:00
Peter Steinberger
874306a2ac test: use synthetic task delivery fixtures 2026-04-21 00:49:06 +01:00
Peter Steinberger
aecd709dfe test: use synthetic task channel fixtures 2026-04-21 00:47:12 +01:00
Peter Steinberger
a8b81fa8e5 fix: guard qqbot channel API fetch 2026-04-21 00:43:50 +01:00
Peter Steinberger
039d1010fe test: fix landing gate helpers 2026-04-21 00:43:50 +01:00
Peter Steinberger
46fdc7d610 docs: update changelog for #68310 2026-04-21 00:43:50 +01:00
Watchtower
dcb525de50 fix(pi-embedded-runner): gate silent-error retry on replay safety
Per @steipete review on #68310: the silent-error retry must not fire when the
failed attempt already recorded potential side effects (messaging tool sent,
cron add, or a mutating tool call that wasn't round-tripped as replay-safe).
Otherwise resubmission can duplicate those actions.

Adds `!attempt.replayMetadata.hadPotentialSideEffects` to the retry condition,
mirroring the gate used by resolveEmptyResponseRetryInstruction and the
planning-only / reasoning-only retry resolvers in run/incomplete-turn.ts.

Adds a new negative regression test:
  "does not retry when the failed attempt recorded side effects"
which reproduces the reviewer's repro — stopReason=error + output=0 + empty
content, but replayMetadata={hadPotentialSideEffects: true, replaySafe: false}.
Expected: no retry, surfaces incomplete-turn error. Confirmed locally.
2026-04-21 00:43:50 +01:00
Watchtower
5fb302ebf1 fix(pi-embedded-runner): retry silent stopReason=error turns (non-frontier models)
ollama/glm-5.1:cloud (and occasionally other models) can end a turn with
stopReason="error", usage.output=0, and empty content[] after a successful
tool-call sequence. The existing empty-response retry path in
src/agents/pi-embedded-runner/run/incomplete-turn.ts is gated on
isStrictAgenticSupportedProviderModel (gpt-5 family only), so non-frontier
models fall through to "incomplete turn detected" with payloads=0 and no
recovery. The user sees no reply and has to nudge.

Add a narrow, model-agnostic resubmission inside the attempt loop, placed
before the incompleteTurnText surface-to-user return:

  - stopReason === "error"
  - usage.output === 0
  - content.length === 0   (excludes reasoning-only error turns)
  - bounded by MAX_EMPTY_ERROR_RETRIES = 3

No instruction injection, no model gating; same prompt, same session
transcript (tool results already captured), just let the loop try again.

New test file run.empty-error-retry.test.ts covers:
  1. Retries for ollama/glm-5.1:cloud → succeeds on 2nd attempt.
  2. Caps at 3 retries → 4 total attempts → surfaces incomplete-turn error.
  3. Does NOT retry when output > 0 (preserve produced text).
  4. Does NOT retry when stopReason=stop + output=0 (NO_REPLY path).
  5. Retries for anthropic/claude-opus-4-7 too — model-agnostic.

Relates to #68281.
2026-04-21 00:43:50 +01:00
Peter Steinberger
982b1c9464 test(ci): reduce channel contract import cost 2026-04-21 00:40:07 +01:00
Peter Steinberger
92191d37e6 test: split chat view coverage 2026-04-21 00:35:58 +01:00
Peter Steinberger
503af7afa6 refactor: dedupe install scan skill spec 2026-04-21 00:32:42 +01:00
Peter Steinberger
1a834a0ff6 test: reuse runtime sidecar uniqueness helper 2026-04-21 00:32:42 +01:00