When a configured Google provider/model row had no explicit
but had a baseUrl set, the fallback defaulted to openai-completions,
causing Gemini requests to route through the OpenAI Responses
transport instead of the native @google/genai transport.
Made resolveConfiguredProviderDefaultApi provider-aware: for the
google provider, the default API is now google-generative-ai.
Root cause: the generic fallback assumed any provider with a baseUrl
should use openai-completions, which is incorrect for Google's native
Gemini API.
Co-authored-by: xin <1052326311+xin@users.noreply.github.com>
Bumps OpenClaw release metadata to 2026.5.31 across package manifests, app version files, plugin metadata, changelog headings, and generated shrinkwraps.
Verification:
- pnpm plugins:sync:check
- pnpm ios:version:check
- pnpm deps:shrinkwrap:check
- git diff --check
- stale 2026.5.30/build-code scan across changed files
- autoreview clean: no accepted/actionable findings
- PR CI green for real gates: Checks, security scans, dependency guard, app lanes, real behavior proof
Known non-code workflow issue:
- label workflow failed because this PR hits GitHub's 100-label issue cap before the size-label step.
* feat: add Claude Opus 4.8 support
* fix: omit Vertex Opus sampling overrides
* fix: preserve Opus adaptive thinking levels
* fix: clamp Anthropic max effort support
* fix: use sha256 for QA mock call ids
* fix: type Anthropic transport test model metadata
* test: update PDF model default for Opus 4.8
Fixes#87191. Keeps Brave and Gemini runtime-injected web search provider config readable by providers without re-exposing legacy tools.web.search provider objects to config validation.
Fix Gemini cached-content GenerateContent payloads so cached requests no longer resend request-level systemInstruction, tools, or toolConfig.
Covers explicit cachedContent and managed cacheRetention prompt caching; fixes#84919.
Proof: Real behavior proof passed on PR head 198a42bbc6 after live Gemini repro/fix evidence was added to the PR body. Focused tests and check:changed were already green.
Thanks @neeravmakwana.
Normalize Google Gemini 3.1 Flash Lite routing to the GA model id and keep the retired preview spelling as a compatibility alias. Align default alias docs, FAQ guidance, and deprecated-model manifest recommendations with the GA id.
Fixes#86151.
Co-authored-by: Sebastien Tardif <sebtardif@ncf.ca>
Fix Google Vertex production ADC mode support by routing explicit google-vertex models to the Vertex transport and relying on google-auth-library for request-time ADC resolution.
Verification:
- pnpm install --frozen-lockfile
- pnpm test extensions/google/transport-stream.test.ts extensions/google/index.test.ts src/config/zod-schema.models.test.ts src/agents/pi-embedded-runner/model.inline-provider.test.ts -- --reporter=verbose
- pnpm check:changed
- GitHub PR checks green on c4b7cad4df
- Live ADC smoke reached Google Vertex auth/transport and failed only because the configured redacted project has the Vertex AI API disabled
Co-authored-by: Damian Finol <damian@felixpago.com>
Summary:
- The branch adds a Google provider thinking-policy resolver and opt-in profile flag, updates shared thinking validation and cron/proof-policy tests, and adjusts ClawSweeper proof parsing.
- Reproducibility: yes. source-reproducible: current main applies the generic off-only profile before provider ... figured thinking through that resolver. I did not execute a live systemd cron run in this read-only review.
Automerge notes:
- PR branch already contained follow-up commit before automerge: fix: preserve Google Gemini 3 cron thinking
Validation:
- ClawSweeper review passed for head a6cd2e826e.
- Required merge gates passed before the squash merge.
Prepared head SHA: a6cd2e826e
Review: https://github.com/openclaw/openclaw/pull/85300#issuecomment-4517662575
Co-authored-by: Neerav Makwana <261249544+neeravmakwana@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.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(gemini): strip sub-second precision from web_search time_range_filter
Gemini's google_search.time_range_filter rejects any non-zero fractional
seconds with "[FIELD_INVALID] Granularity of nano is not supported", even
though the underlying google.protobuf.Timestamp type accepts 0/3/6/9
fractional digits per its public spec. The grounding endpoint enforces a
stricter rule than the underlying type.
Date.prototype.toISOString() always emits millisecond precision, so every
freshness call (and any date_after/date_before call hitting the "now"
fallback for endTime) failed with the above 400 after #66498's fix shipped
in 2026.5.19.
Introduce toGeminiTimeRangeTimestamp() which strips the fractional-second
component before serializing, and route all four timeRangeFilter timestamp
sites through it. isoDateExclusiveEnd happens to produce all-zero
fractional today (so Gemini accepts it), but routing it through the helper
keeps the contract uniform and resilient to future changes.
Why this slipped past the original CI: the existing freshness test used
vi.setSystemTime(new Date("2026-04-15T12:00:00Z")), which always
serializes back as ".000Z" — the one fractional form Gemini happens to
accept. Wall-clock new Date() in production always has non-zero ms. The
new test uses setSystemTime(new Date("2026-04-15T12:00:00.123Z")) to
exercise the realistic case.
Verified empirically against the live Gemini REST API:
".123Z" → 400 "Granularity of nano is not supported"
".000Z" → grounded content (the one fractional form accepted)
"Z" → grounded content
Fixes#85061.
* test(gemini): use realistic non-zero ms in existing freshness test
The original test set the fake clock to a moment with zero fractional
seconds, so toISOString() produced ".000Z" — the one fractional form
Gemini's google_search.time_range_filter happens to accept. Wall-clock
new Date() in production produces non-zero ms, which Gemini rejects.
Bumping the fake time to .123Z makes the existing test exercise the
realistic case alongside the dedicated regression test.
---------
Co-authored-by: Noah R <Noerr@users.noreply.github.com>
