fix: resolve google provider default API to google-generative-ai by 1052326311 · Pull Request #88512 · openclaw/openclaw

1052326311 · 2026-05-31T03:42:41Z

Summary

Keeps the generic baseUrl-only fallback for non-Google providers on openai-completions.
Lets provider transport hooks claim omitted provider defaults before the generic fallback is applied.
Makes the Google provider hook resolve baseUrl-only models.providers.google configs to google-generative-ai, while explicit api: openai-completions stays unchanged.

Verification

OPENCLAW_VITEST_NO_OUTPUT_TIMEOUT_MS=300000 OPENCLAW_VITEST_FS_MODULE_CACHE_PATH=/tmp/pr88512-model-vitest-cache2 fnm exec --using=24.15.0 node scripts/run-vitest.mjs run src/agents/embedded-agent-runner/model.test.ts extensions/google/api.test.ts
fnm exec --using=24.15.0 node scripts/run-tsgo.mjs -p test/tsconfig/tsconfig.test.src.json --incremental --tsBuildInfoFile /tmp/pr88512-test-src2.tsbuildinfo
fnm exec --using=24.15.0 node scripts/run-tsgo.mjs -p test/tsconfig/tsconfig.extensions.test.json --incremental --tsBuildInfoFile /tmp/pr88512-extensions-test.tsbuildinfo
git diff --check

Real behavior proof

Behavior addressed: Google provider configs with models.providers.google.baseUrl and omitted api now resolve to the native Gemini google-generative-ai transport instead of the OpenAI-compatible transport.

Real environment tested: Local OpenClaw checkout at 5a15bd1c79740f7538d31c7cbde2483c4887b0a3, Node 24.15.0, Google Gemini API key read from 1Password item AI API Key - Google Gemini - GEMINI_API_KEY - steipete-m5; key value was not printed.

Exact steps or command run after this patch: Resolved the Google provider transport for provider google, omitted api, and baseUrl https://generativelanguage.googleapis.com, then used the same 1Password Gemini key for a live generateContent request to gemini-2.0-flash.

Evidence after fix: Copied live output from the redacted terminal capture:

head=5a15bd1c79740f7538d31c7cbde2483c4887b0a3
provider=google
configuredApi=omitted
configuredBaseUrl=https://generativelanguage.googleapis.com
resolvedTransport={"api":"google-generative-ai","baseUrl":"https://generativelanguage.googleapis.com/v1beta"}
secretPrefix=AIza
secretSource=1password:AI API Key - Google Gemini - GEMINI_API_KEY - steipete-m5
googleGenerateContent={"status":200,"ok":true,"textLen":3,"containsOk":true}

Observed result after fix: The omitted-api Google config selected the native Gemini transport, and the live Google endpoint accepted the configured key with HTTP 200.

What was not tested: A full packaged gateway chat loop was not run; the focused resolver path, Google provider hook, and live Gemini endpoint were verified directly.

clawsweeper · 2026-05-31T03:44:43Z

Codex review: needs real behavior proof before merge. Reviewed May 30, 2026, 11:49 PM ET / 03:49 UTC.

Summary
The PR changes resolveConfiguredProviderDefaultApi so Google provider configs with a baseUrl and no explicit api default to google-generative-ai, and passes the provider id at its call sites.

PR surface: Source +7. Total +7 across 1 file.

Reproducibility: yes. from source, not from a live run: current main defaults baseUrl-only provider configs to openai-completions, and the linked report shows Google requests being routed through the wrong OpenAI-style transport.

Review metrics: 1 noteworthy metric.

Provider default surface: 1 changed default. The PR changes the omitted-api fallback for baseUrl-only Google provider configs, which affects upgrade and routing behavior beyond a local code path.

Merge readiness
Overall: 🧂 unranked krab
Proof: 🧂 unranked krab
Patch quality: 🦐 gold shrimp
Result: blocked until real behavior proof from a real setup is added.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

Move the provider-specific fallback into Google provider policy or a generic provider-owned hook.
[P1] Add focused regression coverage for baseUrl-only Google config resolving to google-generative-ai while explicit openai-completions remains unchanged.
[P1] Add redacted real behavior proof in the PR body; if re-review does not trigger, ask a maintainer to comment @clawsweeper re-review.

Proof guidance:

[P1] Needs real behavior proof before merge: Only test claims are present; the contributor needs redacted terminal/log/live output showing a real Gemini chat now uses google-generative-ai before merge. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Risk before merge

[P2] Provider routing fallback is compatibility-sensitive; existing models.providers.google rows that intentionally used a Google OpenAI-compatible baseUrl without an explicit api would switch transports unless maintainers require explicit api: openai-completions.
[P1] The PR body reports test runs, but no real Gemini/OpenClaw run, redacted logs, terminal output, or linked artifact demonstrates that a chat request now reaches the native Google transport.
[P2] The diff does not add focused regression coverage for the baseUrl-only Google default; neighboring tests cover explicit api and discovered-api paths but not the new fallback branch.

Maintainer options:

Move the default to provider policy (recommended)
Before merge, keep core fallback generic, have the Google provider resolve the undefined-api/baseUrl case, add focused regression coverage, and attach redacted real behavior proof.
Accept native Google as the bare default
Maintainers can decide that models.providers.google without explicit api always means native Gemini, but should document that OpenAI-compatible Google endpoints must set api: openai-completions.

Next step before merge

[P1] Needs contributor proof and maintainer review of the provider-boundary and compatibility choice; it is not safe for automated repair while proof is missing.

Security
Cleared: The diff only changes provider transport selection logic and does not introduce dependency, workflow, secret-storage, or supply-chain changes.

Review findings

[P2] Move the Google default into provider policy — src/agents/embedded-agent-runner/model.ts:351-352

Review details

Best possible solution:

Move the provider-specific default into the Google provider runtime/policy hook or a generic provider-owned hook contract, add focused regression coverage for baseUrl-only Google config, and require redacted real behavior proof before merge.

Do we have a high-confidence way to reproduce the issue?

Yes from source, not from a live run: current main defaults baseUrl-only provider configs to openai-completions, and the linked report shows Google requests being routed through the wrong OpenAI-style transport.

Is this the best way to solve the issue?

No: the intended behavior is right, but the current patch puts a Google-specific fallback in core instead of the Google provider boundary or a generic provider hook.

Full review comments:

[P2] Move the Google default into provider policy — src/agents/embedded-agent-runner/model.ts:351-352
This core fallback now hard-codes provider === "google", but Google transport/default behavior is owned by the Google provider hooks (normalizeTransport/normalizeConfig). Please resolve the undefined-api/baseUrl case through the provider policy or a generic provider hook so core stays provider-agnostic and future provider aliases do not drift.
Confidence: 0.86

Overall correctness: patch is incorrect
Overall confidence: 0.86

AGENTS.md: found and applied where relevant.

Codex review notes: model gpt-5.5, reasoning high; reviewed against e1a98171417c.

Label changes

Label changes:

add P1: The linked bug breaks Google Gemini chat routing for real model use, but the PR still needs a safe provider-boundary fix and proof before merge.
add merge-risk: 🚨 compatibility: Changing the omitted-api default for existing Google provider configs can alter upgrade behavior for users relying on the prior OpenAI-compatible fallback.
add merge-risk: 🚨 auth-provider: The diff changes provider transport routing for Google models, which directly affects which credential and endpoint family receives the request.
add rating: 🧂 unranked krab: Overall readiness is 🧂 unranked krab; proof is 🧂 unranked krab and patch quality is 🦐 gold shrimp.
add status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs real behavior proof before merge: Only test claims are present; the contributor needs redacted terminal/log/live output showing a real Gemini chat now uses google-generative-ai before merge. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Label justifications:

P1: The linked bug breaks Google Gemini chat routing for real model use, but the PR still needs a safe provider-boundary fix and proof before merge.
merge-risk: 🚨 compatibility: Changing the omitted-api default for existing Google provider configs can alter upgrade behavior for users relying on the prior OpenAI-compatible fallback.
merge-risk: 🚨 auth-provider: The diff changes provider transport routing for Google models, which directly affects which credential and endpoint family receives the request.
rating: 🧂 unranked krab: Overall readiness is 🧂 unranked krab; proof is 🧂 unranked krab and patch quality is 🦐 gold shrimp.
status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs real behavior proof before merge: Only test claims are present; the contributor needs redacted terminal/log/live output showing a real Gemini chat now uses google-generative-ai before merge. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Evidence reviewed

PR surface:

Source +7. Total +7 across 1 file.

View PR surface stats

Area	Files	Added	Removed	Net
Source	1	11	4	+7
Tests	0	0	0	0
Docs	0	0	0	0
Config	0	0	0	0
Generated	0	0	0	0
Other	0	0	0	0
Total	1	11	4	+7

What I checked:

Current fallback: Current main returns an explicit configured api when present, otherwise any baseUrl-only provider config defaults to openai-completions without provider context. (src/agents/embedded-agent-runner/model.ts:340, e1a98171417c)
PR fallback change: The PR head adds a provider parameter, returns google-generative-ai when the provider is exactly google, and passes the provider at the three call sites. (src/agents/embedded-agent-runner/model.ts:340, b14ff960d60c)
Google provider boundary: The bundled Google provider already owns transport normalization and only creates the Google stream function for model.api === "google-generative-ai", so the default can be resolved in provider policy instead of core fallback logic. (extensions/google/provider-registration.ts:47, e1a98171417c)
Explicit OpenAI-compatible path: Existing tests preserve explicit api: "openai-completions" for Google provider rows using the Google /openai endpoint; this PR changes only the omitted-api default. (src/config/model-alias-defaults.test.ts:284, e1a98171417c)
Review policy: Root policy treats provider routing, config/default, and fallback behavior as compatibility-sensitive and asks review to inspect the whole provider/model routing surface. (AGENTS.md:26, e1a98171417c)
Proof gap: The PR changes only model.ts and the supplied GitHub context has test-run claims but no real Gemini/OpenClaw terminal output, logs, screenshot, recording, or linked artifact showing the fixed runtime path. (b14ff960d60c)

Likely related people:

steipete: Current-main blame and log history show Peter Steinberger on the existing provider default helper and recent model-provider refactors in this checkout; older provenance is limited by the available grafted history. (role: recent area contributor; confidence: medium; commits: 22e4289d3f05, 4c33aaa86c16, 00d8d7ead059; files: src/agents/embedded-agent-runner/model.ts, extensions/google/provider-registration.ts, extensions/google/provider-policy.ts)

What the crustacean ranks mean

🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

How this review workflow works

ClawSweeper keeps one durable marker-backed review comment per issue or PR.
Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
Maintainers can also comment @clawsweeper review to request a fresh review only.
Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

byungskers · 2026-05-31T08:02:17Z

This seems to target the right misrouting symptom, but I am a little nervous about hard-coding provider === "google" inside the generic fallback helper. Would it be cleaner to resolve this through Google provider policy / transport normalization instead, so the core fallback stays provider-agnostic and future provider-specific defaults do not accumulate here? Even if you keep the current shape, a focused regression for the baseUrl-only Google case would help lock the intended contract down.

…nclaw#88480) 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.

…nclaw#88480) (openclaw#88512) 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>

1052326311 mentioned this pull request May 31, 2026

Google Gemini chat model routes to openai-responses transport (401), native @google/genai transport never selected #88480

Closed

openclaw-barnacle Bot added agents Agent runtime and tooling size: XS triage: needs-real-behavior-proof Candidate: external PR needs after-fix proof from a real setup. labels May 31, 2026

steipete self-assigned this May 31, 2026

steipete force-pushed the fix/88480-gemini-transport-routing branch from b14ff96 to 5a15bd1 Compare May 31, 2026 21:42

openclaw-barnacle Bot added extensions: google size: S proof: supplied External PR includes structured after-fix real behavior proof. and removed size: XS triage: needs-real-behavior-proof Candidate: external PR needs after-fix proof from a real setup. labels May 31, 2026

steipete merged commit b73e135 into openclaw:main May 31, 2026
199 of 207 checks passed

amitV mentioned this pull request Jun 1, 2026

[Feature]: updating openclaw to use gemini-3.1-flash-lite instead of gemini-3.1-flash-lite-preview #80380

Open

koverholt mentioned this pull request Jun 1, 2026

fix(google-vertex): add Google Vertex AI onboarding wizard, fix ADC auth, add provider docs #87800

Open

paulogogs mentioned this pull request Jun 5, 2026

[Bug]: [Bug]: google-vertex models fail with model_not_found at runtime on 2026.5.28 and 2026.6.1 — direct Vertex API calls succeed with same credentials #90506

Closed

syfvb mentioned this pull request Jun 10, 2026

[Performance] /models command slow in v2026.6.1 — catalog loading regression #91809

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix: resolve google provider default API to google-generative-ai#88512

fix: resolve google provider default API to google-generative-ai#88512
steipete merged 1 commit into
openclaw:mainfrom
1052326311:fix/88480-gemini-transport-routing

1052326311 commented May 31, 2026 •

edited by steipete

Loading

Uh oh!

clawsweeper Bot commented May 31, 2026 •

edited

Loading

Uh oh!

byungskers commented May 31, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Conversation

1052326311 commented May 31, 2026 • edited by steipete Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Verification

Real behavior proof

Uh oh!

clawsweeper Bot commented May 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

byungskers commented May 31, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

1052326311 commented May 31, 2026 •

edited by steipete

Loading

clawsweeper Bot commented May 31, 2026 •

edited

Loading