fix(google): resolve Gemini 3.1 models for all Google provider aliases by hydro13 · Pull Request #56567 · openclaw/openclaw

hydro13 · 2026-03-28T18:51:31Z

Summary

The forward-compat resolver hardcoded "google" as the provider ID for template lookup, so alias providers (google-vertex, google-gemini-cli) could not find matching templates
Pass actual provider ID from runtime context + add templateProviderId fallback for cross-provider template resolution
Fix flash-lite prefix ordering: check gemini-3.1-flash-lite before gemini-3.1-flash to prevent misclassification
Add regression tests for pro, flash, and flash-lite across provider aliases

Root Cause

resolveGoogle31ForwardCompatModel() in extensions/google/provider-models.ts was called with hardcoded providerId: "google" from extensions/google/index.ts:243. When users configured google-vertex or other aliases, the template lookup failed because no templates existed under that provider ID.

Secondary bug: gemini-3.1-flash-lite starts with gemini-3.1-flash, so without explicit ordering it matched the wrong branch.

Change Type

Bug fix

Testing

24 tests pass across 3 test files including new regression coverage for:

Gemini 3.1 Pro via google-vertex alias
Gemini 3.1 Flash via direct google provider
Gemini 3.1 Flash-Lite prefix ordering

Fixes #36111

greptile-apps · 2026-03-28T18:54:38Z

Greptile Summary

This PR fixes resolveGoogle31ForwardCompatModel so that Google provider aliases (google-vertex, google-antigravity) can resolve Gemini 3.1 models by falling back to google-gemini-cli templates, rather than failing silently because of a hardcoded "google" provider ID. It also corrects a prefix-ordering bug where gemini-3.1-flash-lite was mis-classified as gemini-3.1-flash.

Root cause fix (index.ts): Replaces providerId: \"google\" with providerId: ctx.provider and adds templateProviderId: GOOGLE_GEMINI_CLI_PROVIDER_ID as a fallback, so alias providers reach the correct template source.
New helper (provider-models.ts): cloneFirstGoogleTemplateModel iterates candidate provider IDs (primary → fallback) and deduplicates them via Set, which is clean and correct.
Prefix ordering fix (provider-models.ts): GEMINI_3_1_FLASH_LITE_PREFIX is now checked before GEMINI_3_1_FLASH_PREFIX, preventing misclassification.
Flash-lite template list (provider-models.ts): GEMINI_3_1_FLASH_LITE_TEMPLATE_IDS falls back to \"gemini-3-flash-preview\" when no flash-lite-specific template is registered. Flash and flash-lite have different capabilities, so this silent fallback could propagate incorrect metadata (context window, pricing). This path is also not covered by the new tests.
google-gemini-cli provider (index.ts:141–142): Left unchanged — its own resolveDynamicModel still calls the function without templateProviderId. This is consistent with it being a standalone provider, but a comment clarifying the intentional omission would help future readers.

Confidence Score: 5/5

Safe to merge; the fix is well-scoped and the two remaining findings are P2 style/coverage concerns that do not block correctness in production.

All identified issues are P2: the flash-lite → flash fallback is a potential metadata inaccuracy without an immediate reproducer, and the google-gemini-cli comment gap is purely cosmetic. The primary regression (alias providers returning undefined for Gemini 3.1 models) is fixed and covered by new tests. No P0/P1 issues found.

extensions/google/provider-models.ts — review GEMINI_3_1_FLASH_LITE_TEMPLATE_IDS fallback to "gemini-3-flash-preview"

Important Files Changed

Filename	Overview
extensions/google/provider-models.ts	Adds flash-lite prefix/template handling and a new cloneFirstGoogleTemplateModel helper that iterates provider IDs (primary then fallback) for template resolution. Prefix ordering fix (flash-lite before flash) is correct. Minor concern: flash-lite falls back to a flash template when its own template is absent.
extensions/google/index.ts	Replaces hardcoded "google" with ctx.provider and adds templateProviderId fallback for the main google provider registration. Correctly resolves templates for hook aliases (google-vertex, google-antigravity). google-gemini-cli provider registration left unchanged (intentional).
extensions/google/provider-models.test.ts	New test file covering pro via google-vertex alias, flash via direct google provider, and flash-lite prefix ordering. Covers the primary regression scenarios. Missing coverage for: flash-lite falling back to flash template, and google-antigravity alias.

Comments Outside Diff (1)

extensions/google/index.ts, line 141-142 (link)

google-gemini-cli provider missing templateProviderId after refactor

The google provider registration was updated to pass templateProviderId: GOOGLE_GEMINI_CLI_PROVIDER_ID so that alias providers can fall back to google-gemini-cli templates. The google-gemini-cli provider's own resolveDynamicModel was not touched and still calls resolveGoogle31ForwardCompatModel without templateProviderId.

This is consistent with google-gemini-cli being a standalone provider (not an alias), and the current tests pass because its templates are registered under its own ID. However, if google-gemini-cli ever needs a cross-provider template fallback in the future, no fallback mechanism exists here. A brief comment explaining this intentional omission would help future readers.

Prompt To Fix With AI

This is a comment left during a code review.
Path: extensions/google/index.ts
Line: 141-142

Comment:
**`google-gemini-cli` provider missing `templateProviderId` after refactor**

The `google` provider registration was updated to pass `templateProviderId: GOOGLE_GEMINI_CLI_PROVIDER_ID` so that alias providers can fall back to `google-gemini-cli` templates. The `google-gemini-cli` provider's own `resolveDynamicModel` was not touched and still calls `resolveGoogle31ForwardCompatModel` without `templateProviderId`.

This is consistent with `google-gemini-cli` being a standalone provider (not an alias), and the current tests pass because its templates are registered under its own ID. However, if `google-gemini-cli` ever needs a cross-provider template fallback in the future, no fallback mechanism exists here. A brief comment explaining this intentional omission would help future readers.

How can I resolve this? If you propose a fix, please make it concise.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

Prompt To Fix All With AI

This is a comment left during a code review.
Path: extensions/google/provider-models.ts
Line: 11-14

Comment:
**Flash-lite falls back to flash template metadata**

`GEMINI_3_1_FLASH_LITE_TEMPLATE_IDS` includes `"gemini-3-flash-preview"` as a fallback when `"gemini-3.1-flash-lite-preview"` isn't found in the registry. Flash and flash-lite have different capabilities (context window, pricing, token limits), so cloning flash metadata for a flash-lite model request would silently produce an incorrect model descriptor.

None of the new tests cover this fallback path (i.e. the case where only `"gemini-3-flash-preview"` is registered but `"gemini-3.1-flash-lite-preview"` is requested). Consider either:
1. Removing the `"gemini-3-flash-preview"` fallback from `GEMINI_3_1_FLASH_LITE_TEMPLATE_IDS` and returning `undefined` when the dedicated template is absent, or
2. Adding a test asserting the expected behaviour when the fallback is exercised, and documenting why borrowing flash properties for flash-lite is acceptable.

How can I resolve this? If you propose a fix, please make it concise.

---

This is a comment left during a code review.
Path: extensions/google/index.ts
Line: 141-142

Comment:
**`google-gemini-cli` provider missing `templateProviderId` after refactor**

The `google` provider registration was updated to pass `templateProviderId: GOOGLE_GEMINI_CLI_PROVIDER_ID` so that alias providers can fall back to `google-gemini-cli` templates. The `google-gemini-cli` provider's own `resolveDynamicModel` was not touched and still calls `resolveGoogle31ForwardCompatModel` without `templateProviderId`.

This is consistent with `google-gemini-cli` being a standalone provider (not an alias), and the current tests pass because its templates are registered under its own ID. However, if `google-gemini-cli` ever needs a cross-provider template fallback in the future, no fallback mechanism exists here. A brief comment explaining this intentional omission would help future readers.

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (1): Last reviewed commit: "fix(google): resolve Gemini 3.1 models f..." | Re-trigger Greptile}

greptile-apps · 2026-03-28T18:54:42Z

+const GEMINI_3_1_FLASH_LITE_TEMPLATE_IDS = [
+  "gemini-3.1-flash-lite-preview",
+  "gemini-3-flash-preview",
+] as const;


Flash-lite falls back to flash template metadata

GEMINI_3_1_FLASH_LITE_TEMPLATE_IDS includes "gemini-3-flash-preview" as a fallback when "gemini-3.1-flash-lite-preview" isn't found in the registry. Flash and flash-lite have different capabilities (context window, pricing, token limits), so cloning flash metadata for a flash-lite model request would silently produce an incorrect model descriptor.

None of the new tests cover this fallback path (i.e. the case where only "gemini-3-flash-preview" is registered but "gemini-3.1-flash-lite-preview" is requested). Consider either:

Removing the "gemini-3-flash-preview" fallback from GEMINI_3_1_FLASH_LITE_TEMPLATE_IDS and returning undefined when the dedicated template is absent, or

Adding a test asserting the expected behaviour when the fallback is exercised, and documenting why borrowing flash properties for flash-lite is acceptable.

Prompt To Fix With AI

This is a comment left during a code review. Path: extensions/google/provider-models.ts Line: 11-14 Comment: **Flash-lite falls back to flash template metadata** `GEMINI_3_1_FLASH_LITE_TEMPLATE_IDS` includes `"gemini-3-flash-preview"` as a fallback when `"gemini-3.1-flash-lite-preview"` isn't found in the registry. Flash and flash-lite have different capabilities (context window, pricing, token limits), so cloning flash metadata for a flash-lite model request would silently produce an incorrect model descriptor. None of the new tests cover this fallback path (i.e. the case where only `"gemini-3-flash-preview"` is registered but `"gemini-3.1-flash-lite-preview"` is requested). Consider either: 1. Removing the `"gemini-3-flash-preview"` fallback from `GEMINI_3_1_FLASH_LITE_TEMPLATE_IDS` and returning `undefined` when the dedicated template is absent, or 2. Adding a test asserting the expected behaviour when the fallback is exercised, and documenting why borrowing flash properties for flash-lite is acceptable. How can I resolve this? If you propose a fix, please make it concise.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 326a068cc0

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

The forward-compat resolver hardcoded 'google' as the provider ID for template lookup, so alias providers (google-vertex, google-gemini-cli) could not find matching templates. Pass the actual provider ID from the runtime context and add a templateProviderId fallback for cross-provider template resolution. Also fix flash-lite prefix ordering — check 'gemini-3.1-flash-lite' before 'gemini-3.1-flash' to prevent misclassification. Add regression tests for pro, flash, and flash-lite across provider aliases. Fixes openclaw#36111

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 689b9c7f53

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

chatgpt-codex-connector · 2026-03-28T19:06:40Z

+      patch: {
+        ...params.patch,
+        provider: params.providerId,
+      },


Preserve transport fields when cloning cross-provider templates

When this helper clones from templateProviderId, it rewrites only provider and keeps the template api/baseUrl, so alias resolution can now return mixed models (fresh evidence: the new test asserts provider: "google-vertex" with api: "google-gemini-cli"). That breaks downstream assumptions keyed on provider ID (for example sanitizeToolsForGoogle only runs for google-gemini-cli), so tool schemas are left unsanitized and Cloud Code requests can fail at runtime on unsupported JSON Schema keywords.

Useful? React with 👍 / 👎.

Add missing changelog entries for PRs openclaw#56500, openclaw#56540, openclaw#56555, openclaw#56567, openclaw#56573, openclaw#56587, openclaw#56595, openclaw#56612, openclaw#56620.

Add missing changelog entries for PRs #56500, #56540, #56555, #56567, #56573, #56587, #56595, #56612, #56620.

openclaw#56567) The forward-compat resolver hardcoded 'google' as the provider ID for template lookup, so alias providers (google-vertex, google-gemini-cli) could not find matching templates. Pass the actual provider ID from the runtime context and add a templateProviderId fallback for cross-provider template resolution. Also fix flash-lite prefix ordering — check 'gemini-3.1-flash-lite' before 'gemini-3.1-flash' to prevent misclassification. Add regression tests for pro, flash, and flash-lite across provider aliases. Fixes openclaw#36111

Add missing changelog entries for PRs openclaw#56500, openclaw#56540, openclaw#56555, openclaw#56567, openclaw#56573, openclaw#56587, openclaw#56595, openclaw#56612, openclaw#56620.

openclaw#56567) The forward-compat resolver hardcoded 'google' as the provider ID for template lookup, so alias providers (google-vertex, google-gemini-cli) could not find matching templates. Pass the actual provider ID from the runtime context and add a templateProviderId fallback for cross-provider template resolution. Also fix flash-lite prefix ordering — check 'gemini-3.1-flash-lite' before 'gemini-3.1-flash' to prevent misclassification. Add regression tests for pro, flash, and flash-lite across provider aliases. Fixes openclaw#36111

Add missing changelog entries for PRs openclaw#56500, openclaw#56540, openclaw#56555, openclaw#56567, openclaw#56573, openclaw#56587, openclaw#56595, openclaw#56612, openclaw#56620.

openclaw#56567) The forward-compat resolver hardcoded 'google' as the provider ID for template lookup, so alias providers (google-vertex, google-gemini-cli) could not find matching templates. Pass the actual provider ID from the runtime context and add a templateProviderId fallback for cross-provider template resolution. Also fix flash-lite prefix ordering — check 'gemini-3.1-flash-lite' before 'gemini-3.1-flash' to prevent misclassification. Add regression tests for pro, flash, and flash-lite across provider aliases. Fixes openclaw#36111

Add missing changelog entries for PRs openclaw#56500, openclaw#56540, openclaw#56555, openclaw#56567, openclaw#56573, openclaw#56587, openclaw#56595, openclaw#56612, openclaw#56620.

openclaw#56567) The forward-compat resolver hardcoded 'google' as the provider ID for template lookup, so alias providers (google-vertex, google-gemini-cli) could not find matching templates. Pass the actual provider ID from the runtime context and add a templateProviderId fallback for cross-provider template resolution. Also fix flash-lite prefix ordering — check 'gemini-3.1-flash-lite' before 'gemini-3.1-flash' to prevent misclassification. Add regression tests for pro, flash, and flash-lite across provider aliases. Fixes openclaw#36111

Add missing changelog entries for PRs openclaw#56500, openclaw#56540, openclaw#56555, openclaw#56567, openclaw#56573, openclaw#56587, openclaw#56595, openclaw#56612, openclaw#56620.

openclaw#56567) The forward-compat resolver hardcoded 'google' as the provider ID for template lookup, so alias providers (google-vertex, google-gemini-cli) could not find matching templates. Pass the actual provider ID from the runtime context and add a templateProviderId fallback for cross-provider template resolution. Also fix flash-lite prefix ordering — check 'gemini-3.1-flash-lite' before 'gemini-3.1-flash' to prevent misclassification. Add regression tests for pro, flash, and flash-lite across provider aliases. Fixes openclaw#36111

Add missing changelog entries for PRs openclaw#56500, openclaw#56540, openclaw#56555, openclaw#56567, openclaw#56573, openclaw#56587, openclaw#56595, openclaw#56612, openclaw#56620.

openclaw#56567) The forward-compat resolver hardcoded 'google' as the provider ID for template lookup, so alias providers (google-vertex, google-gemini-cli) could not find matching templates. Pass the actual provider ID from the runtime context and add a templateProviderId fallback for cross-provider template resolution. Also fix flash-lite prefix ordering — check 'gemini-3.1-flash-lite' before 'gemini-3.1-flash' to prevent misclassification. Add regression tests for pro, flash, and flash-lite across provider aliases. Fixes openclaw#36111

Add missing changelog entries for PRs openclaw#56500, openclaw#56540, openclaw#56555, openclaw#56567, openclaw#56573, openclaw#56587, openclaw#56595, openclaw#56612, openclaw#56620.

openclaw#56567) The forward-compat resolver hardcoded 'google' as the provider ID for template lookup, so alias providers (google-vertex, google-gemini-cli) could not find matching templates. Pass the actual provider ID from the runtime context and add a templateProviderId fallback for cross-provider template resolution. Also fix flash-lite prefix ordering — check 'gemini-3.1-flash-lite' before 'gemini-3.1-flash' to prevent misclassification. Add regression tests for pro, flash, and flash-lite across provider aliases. Fixes openclaw#36111

Add missing changelog entries for PRs openclaw#56500, openclaw#56540, openclaw#56555, openclaw#56567, openclaw#56573, openclaw#56587, openclaw#56595, openclaw#56612, openclaw#56620.

openclaw-barnacle Bot added size: S maintainer Maintainer-authored PR labels Mar 28, 2026

greptile-apps Bot reviewed Mar 28, 2026

View reviewed changes

chatgpt-codex-connector Bot reviewed Mar 28, 2026

View reviewed changes

Comment thread extensions/google/provider-models.ts

hydro13 force-pushed the fix/gemini-31-provider-routing branch from 326a068 to 689b9c7 Compare March 28, 2026 18:58

hydro13 merged commit d1b0f8e into openclaw:main Mar 28, 2026
8 checks passed

chatgpt-codex-connector Bot reviewed Mar 28, 2026

View reviewed changes

github-actions Bot mentioned this pull request Mar 28, 2026

📡 Upstream Digest — 2026-03-28 20:21 UTC curtismercier/openclaw-mods#394

Open

hydro13 mentioned this pull request Mar 28, 2026

chore: backfill changelog entries for recent fixes #56625

Merged

hydro13 added a commit that referenced this pull request Mar 28, 2026

chore: backfill changelog entries for recent fixes (#56625)

2022dfd

Add missing changelog entries for PRs #56500, #56540, #56555, #56567, #56573, #56587, #56595, #56612, #56620.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(google): resolve Gemini 3.1 models for all Google provider aliases#56567

fix(google): resolve Gemini 3.1 models for all Google provider aliases#56567
hydro13 merged 1 commit intoopenclaw:mainfrom
hydro13:fix/gemini-31-provider-routing

hydro13 commented Mar 28, 2026

Uh oh!

greptile-apps Bot commented Mar 28, 2026 •

edited

Loading

Important Files Changed

Comments Outside Diff (1)

Uh oh!

greptile-apps Bot Mar 28, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Mar 28, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

hydro13 commented Mar 28, 2026

Summary

Root Cause

Change Type

Testing

Uh oh!

greptile-apps Bot commented Mar 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 5/5

Important Files Changed

Comments Outside Diff (1)

Uh oh!

greptile-apps Bot Mar 28, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Mar 28, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

greptile-apps Bot commented Mar 28, 2026 •

edited

Loading