Forward configured stop sequences to Gemini generationConfig.stopSequences in the bundled Google transport, matching the shared Google provider behavior and the @google/genai request contract.\n\nThanks @coder999999999.
(cherry picked from commit 4bb86877e2)
Enable Mistral prompt cache keys without long-retention forwarding. Update cached-read pricing and doctor migration for existing Mistral provider config. Fixes#83709.
Move QQBot credential backups, gateway sessions, known-user records, and ref-index rows into plugin SQLite KV stores. Import shipped JSON/JSONL state files on first use and keep auxiliary known-user/ref-index state best-effort so message delivery is not blocked by cache persistence failures.
* fix(providers): use native reasoning mode for direct Gemini API, keep CLI tagged
Gemini 2.5+ delivers reasoning via native thinkingParts (thinkingConfig.
includeThoughts). Having tagged mode active at the same time injects a
<think>…</think>/<final>…</final> directive into the system prompt; the
model opens a <think> block before a tool call, never closes it, and
returns an empty post-tool turn (content:[], payloads=0 error, #69220).
Fix: override resolveReasoningOutputMode in buildGoogleProvider() only —
not in the shared GOOGLE_GEMINI_PROVIDER_HOOKS. The Gemini CLI backend
(google-gemini-cli) runs gemini --output-format json and parses a text
response field, not native thought parts; it must stay on tagged mode.
A regression test confirms google-gemini-cli remains "tagged".
Also remove the dead BUILTIN_REASONING_OUTPUT_MODES entry keyed on
"google-generative-ai" from provider-utils.ts — that string is only
ever the transport model.api value, never the provider id passed to
resolveReasoningOutputMode, so the map was unreachable.
Fixes#69220
* docs: clarify Gemini reasoning output modes
* fix(google): keep Antigravity reasoning tagged
* fix(google): default direct reasoning checks to native
* fix(google): import reasoning context from plugin entry
---------
Co-authored-by: Peter Steinberger <steipete@gmail.com>
Fixes#88355.
When a resumed Codex context-engine thread overflows and OpenClaw retries on a fresh native thread, clear the stale thread-bootstrap projection metadata from the fresh binding. This prevents later turns from treating that fresh thread as already projected when it only received the bare retry prompt.
Verification:
- Autoreview clean: no accepted/actionable findings reported.
- CI run 26717883204 green on head 5438f8ad34.
* fix(google): add gemini-3.1-flash-lite to provider catalog
Adds the missing gemini-3.1-flash-lite model definition to the
GOOGLE_GEMINI_TEXT_MODELS array. This resolves the ProviderFailoverError
when configuring google-vertex/gemini-3.1-flash-lite.
Fixes#89390
* test(google): cover Gemini flash lite catalog row
---------
Co-authored-by: Peter Steinberger <steipete@gmail.com>
QQBot credential backups now resolve under the active OpenClaw state directory instead of the old home-global QQBot data path. This keeps isolated gateway profiles from restoring each other's QQBot appId/clientSecret backups while preserving per-state-root recovery.
Proof: focused QQBot path/storage-laziness Vitest suite passed on Node 24.15.0, focused oxlint passed, source-runtime two-root backup proof passed, exact-head CI run 26814565282 passed, and ClawSweeper re-review run 26815054980 marked proof sufficient.
Closes#84313.
Co-authored-by: Coy Geek <65363919+coygeek@users.noreply.github.com>
Fix Discord progress-mode reasoning streams so delta chunks accumulate before display formatting, preserving raw Thinking/Reasoning-prefixed content and balanced truncation.\n\nFixes #83983.\n\nThanks @giodl73-repo for the fix and live Discord proof.
Summary:
- The branch adds a memory-core `startup_retry` reconciliation mode and regression tests for quiet startup retries, retry-window exhaustion, and live-config retry semantics.
- PR surface: Source +9, Tests +114. Total +123 across 2 files.
- Reproducibility: yes. from source: current main routes the first startup retry through runtime reconciliatio ... st expects the warn-level `cron service unavailable` log. I did not execute tests in this read-only review.
Automerge notes:
- Ran the ClawSweeper repair loop before final review.
- Included post-review commit in the final squash: fix(memory-core): keep startup cron retries quiet
Validation:
- ClawSweeper review passed for head 7220f940d0.
- Required merge gates passed before the squash merge.
Prepared head SHA: 7220f940d0
Review: https://github.com/openclaw/openclaw/pull/89075#issuecomment-4592446250
Co-authored-by: bennewell35 <newelljben@gmail.com>
Co-authored-by: clawsweeper <274271284+clawsweeper[bot]@users.noreply.github.com>
Co-authored-by: clawsweeper[bot] <274271284+clawsweeper[bot]@users.noreply.github.com>
Approved-by: takhoffman
Co-authored-by: takhoffman <781889+takhoffman@users.noreply.github.com>
Add `ownsNativeCompaction` capability to CliBackendPlugin so backends
that manage their own transcript compaction (e.g. Claude Code) can
declare it once and OpenClaw defers instead of fighting or failing.
Today only Codex declares compaction ownership (via the embedded runner
path + agentHarnessId). Claude-cli never reaches that path because it
runs as a CLI subprocess with no harness id set, so the safeguard
summarizer fires and hard-fails the turn.
This PR:
- Adds `ownsNativeCompaction?: boolean` to the backend plugin type
- Propagates it through all 4 backend resolution paths
- In `runCliTurnCompactionLifecycle`, when a backend declares ownership
but has no harness endpoint, returns a no-op instead of falling
through to the safeguard
- Sets the flag on claude-cli (first adopter)
Codex's existing native-harness path is unchanged: when
`isNativeHarnessCompactionSession` matches, the harness compaction
endpoint is still called as before.
Generalizes the partial fix in #87785 (codex-scoped) to a capability
any backend can opt into.
* fix(qqbot): allow RFC2544 benchmark range for token fetch (#88984)
QQ Bot `bots.qq.com` token-fetch path was failing for users whose DNS resolver maps the hostname into the RFC 2544 benchmark range `198.18.0.0/15` (commonly seen with fake-IP proxy stacks: sing-box, Clash, Surge, WSL2 DNS). The default SSRF guard treats that range as private and blocks the request, surfacing as "Network error getting access_token: Blocked: resolves to private/internal/special-use IP address".
Pass a host-scoped `SsrFPolicy` (`allowRfc2544BenchmarkRange: true`) to the single hard-coded `TOKEN_URL` request, mirroring the existing `QQBOT_MEDIA_SSRF_POLICY` pattern used by the media path. Because `TOKEN_URL` is a const and not user-controlled, the relaxation cannot widen attack surface to other hosts.
Adds a regression test asserting `policy: { allowRfc2544BenchmarkRange: true }` is forwarded into `fetchWithSsrFGuard`, and updates the existing equality assertion accordingly.
Fixes#88984
* fix(qqbot): scope token ssrf policy
Repairs a batch of narrow model/provider edge cases:
- honor OpenAI and Anthropic base URL environment overrides when provider config does not set an explicit base URL
- preserve OpenRouter Anthropic cache retention while stripping unsupported transport options
- allow apply_patch for non-OpenAI providers when the tool config otherwise permits it
- prune stale same-provider model selections from configure/model picker state
- expose GitHub Copilot bundled thinking policy metadata to offline/provider-policy lookups
- repair additive SQLite shared-state upgrades for existing databases
- keep same-size rotated log readers from reusing stale content in CI tooling
Proof:
- GitHub PR checks green on exact head 46514909b0
- Crabbox delegated Blacksmith Testbox tbx_01kt3em5r9vd7g0bnykrff6jdk exited 0
- Focused local Vitest/oxlint/format proof recorded in PR body and land-ready comment
Fixes#80347.
Fixes#88357.
Fixes#45269.
Supersedes #74427, #74432, #79370, #79894, #80366, and #88359.
Summary:
- Adds defensive failed-count reads in auto-reply/ACP accounting and Feishu fallback paths, plus a focused regression test, while keeping `ReplyDispatcher.getFailedCounts` required.
- PR surface: Source +24, Tests +35. Total +59 across 5 files.
- Reproducibility: yes. from source inspection. Current main calls `dispatcher.getFailedCounts().final` and si ... issing that method follows a clear TypeError path; the source PR also supplied terminal before/after proof.
Automerge notes:
- PR branch already contained follow-up commit before automerge: fix(auto-reply): guard missing dispatcher getFailedCounts without wea…
Validation:
- ClawSweeper review passed for head 0bdfb4adeb.
- Required merge gates passed before the squash merge.
Prepared head SHA: 0bdfb4adeb
Review: https://github.com/openclaw/openclaw/pull/89318#issuecomment-4598624344
Co-authored-by: Alix-007 <li.long15@xydigit.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: clawsweeper <274271284+clawsweeper[bot]@users.noreply.github.com>
Co-authored-by: clawsweeper[bot] <274271284+clawsweeper[bot]@users.noreply.github.com>
Approved-by: takhoffman
Co-authored-by: takhoffman <781889+takhoffman@users.noreply.github.com>
Fix live model inference edge cases across provider streaming, model switching, outbound delivery, and gateway tool resolution.
Includes live/provider issue fixes and leaves #89100 explicitly partial for the remaining FM-2 group routing case.
Keep iMessage native typing indicators alive through long tool-running gaps by bridging tool-start activity into the existing typing controller, while preserving typingMode and sendPolicy suppression semantics.
Real behavior proof was added from the live iMessage generated-image run: inbound id 5805, outgoing media reply id 5806, and requester-observed typing during the 84s tool path.
Co-authored-by: omarshahine <10343873+omarshahine@users.noreply.github.com>
Reviewed-by: @lobster
Route Slack plugin approval delivery through the shared native approval route gates while preserving Slack Block Kit buttons and plugin resolver semantics.
Verification: Slack/native approval unit tests, Slack QA Lab, and live clawd native plugin approval via Slack desktop.
