feat(models): declare reasoning efforts for gemini

[AmineAce]

Summary

Declares reasoningEfforts for the Gemini 3.x mappings that were missing them, per #3028.

Changes

• google-vertex: gemini-3.1-flash-lite, gemini-3.5-flash, gemini-3.5-flash-lite, gemini-3.6-flash → ["minimal", "low", "medium", "high"]
• google-ai-studio: gemini-3.1-flash-lite → ["minimal", "low", "medium", "high"] (its Vertex sibling declared the tiers but the AI Studio mapping did not, unlike the other three models which declare them on both providers)

Source: Google Cloud's Vertex Thinking docs — summary table lists exact supported thinking_level values per model, cross-checked against each model's individual Vertex docs page.

Note that the gateway does not forward thinking_level itself; for Google providers it translates each effort tier into a thinkingConfig.thinkingBudget value (prepare-request-body.ts). The declared tiers are what the catalogue publishes and what the e2e suite exercises.

Not changed — with reasons

• Gemini 2.5-era and 1.5-era mappings — Vertex docs confirm no thinking_level mechanism exists for these generations (thinking_budget, a raw token count, is used instead — already declared via reasoningMaxTokens).
• gemini-3-pro-preview, gemini-3.1-flash-lite-preview — retired/deactivated on Vertex.
• gemini-3-flash-preview, gemini-3.1-pro-preview — the Vertex thinking table has rows for "Gemini 3 Flash"/"Gemini 3.1 Pro" without "preview" in the name; couldn't confirm these are the same model as our catalog's preview-suffixed entries (dedicated per-model doc pages both 404). Left unset rather than assume.
• none — not declared anywhere here, matching the sibling mappings. It does return 200 on AI Studio with thinking disabled, but declaring it is out of scope for this change.

Testing

• pnpm format, pnpm build — clean.
• Scoped e2e against live Google endpoints (FULL_MODE=true, which expands one case per declared tier):
  • google-vertex × 4 models: 16/16 effort cases pass (minimal/low/medium/high).
  • google-ai-studio/gemini-3.1-flash-lite: 4/4 effort cases pass; full scoped run 113 passed / 0 failed.
• Verified directly through the gateway that each tier reaches upstream and is differentiated rather than collapsed — reasoning-token counts scale with effort (e.g. gemini-3.1-flash-lite on Vertex: 140 → 152 → 243 → 271 for minimal → low → medium → high), matching Vertex's own thoughtsTokenCount in usageMetadata.

Summary by CodeRabbit

• New Features
  • Added configurable reasoning effort levels—minimal, low, medium, and high—for supported Gemini models across Google AI Studio and Google Vertex AI.
  • Supported models include Gemini 3.1 Flash Lite, Gemini 3.5 Flash, Gemini 3.5 Flash Lite, and Gemini 3.6 Flash.

[coderabbitai]

Walkthrough

Five Google AI Studio and Google Vertex Gemini configurations now declare support for minimal, low, medium, and high reasoning effort levels.

Changes

Google Gemini reasoning effort support

Layer / File(s)	Summary
Gemini reasoning effort declarations packages/models/src/models/google.ts	Adds reasoningEfforts with minimal, low, medium, and high to five Gemini configurations across Google AI Studio and Google Vertex.

Estimated code review effort: 1 (Trivial) | ~2 minutes

Possibly related PRs

• theopenco/llmgateway#3164: Updates configurations for the same Gemini models.
• theopenco/llmgateway#3165: Modifies Gemini model definitions in the same file.
• theopenco/llmgateway#3072: Adds reasoningEfforts declarations to model provider configurations.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly identifies the main change: declaring reasoning efforts for Gemini models.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}