feat: add thinkingLevel and thinkingBudget config for Gemini models

[brendan-kellam]

Summary

• Adds thinkingLevel and thinkingBudget parameters to the GoogleGenerativeAILanguageModel and GoogleVertexLanguageModel schema definitions, allowing per-model thinking configuration.
• thinkingLevel (enum: minimal, low, medium, high) controls reasoning depth for Gemini 3 models. thinkingBudget (integer, -1 for dynamic) sets the thinking token budget for Gemini 2.5 models.
• Wires the new config through to the AI SDK provider options for both google-generative-ai and google-vertex providers.
• Deprecates the GOOGLE_VERTEX_THINKING_BUDGET_TOKENS env var in favor of the new per-model config, with env var fallback for backward compatibility.

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features

  • Added thinkingLevel (minimal|low|medium|high) and thinkingBudget (integer, -1 = dynamic) model config options for Google Generative AI and Google Vertex providers.

• Breaking Changes

  • Removed GOOGLE_VERTEX_INCLUDE_THOUGHTS — thoughts are always included.
  • Deprecated GOOGLE_VERTEX_THINKING_BUDGET_TOKENS in favor of per-model thinkingBudget.

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
sourcebot	🟢 Ready	View Preview	Apr 13, 2026, 10:46 PM

💡 Tip: Enable Workflows to automatically generate PRs for you.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 2f360e6c-cac6-4e09-a9dc-2f428faa03d7

📥 Commits

Reviewing files that changed from the base of the PR and between 3efe2ac and 8204819.

📒 Files selected for processing (1)

• docs/docs/configuration/language-model-providers.mdx

---

Walkthrough

Adds per-model Google thinking options (thinkingLevel, thinkingBudget), updates schemas and TypeScript types, makes GOOGLE_VERTEX_THINKING_BUDGET_TOKENS optional/deprecated and removes GOOGLE_VERTEX_INCLUDE_THOUGHTS, and updates chat runtime to source thinking config from model config (with env fallback for vertex).

Changes

Cohort / File(s)	Summary
Changelog & Docs CHANGELOG.md, docs/docs/configuration/language-model-providers.mdx, docs/snippets/schemas/v3/*	Documented new thinkingLevel and thinkingBudget options for Google providers; noted deprecation/removal of related env vars; added examples.
Schema sources packages/schemas/src/v3/index.schema.ts, packages/schemas/src/v3/languageModel.schema.ts, docs/snippets/schemas/v3/index.schema.mdx, docs/snippets/schemas/v3/languageModel.schema.mdx, schemas/v3/languageModel.json	Added optional thinkingLevel (enum: `minimal
Type definitions packages/schemas/src/v3/index.type.ts, packages/schemas/src/v3/languageModel.type.ts	Added optional properties `thinkingLevel?: "minimal"
Environment schema packages/shared/src/env.server.ts	Made GOOGLE_VERTEX_THINKING_BUDGET_TOKENS optional and marked deprecated; removed GOOGLE_VERTEX_INCLUDE_THOUGHTS from env validation.
Runtime/chat logic packages/web/src/features/chat/utils.server.ts	Populates providerOptions.thinkingConfig from per-model config (thinkingBudget, thinkingLevel) for Google providers; sets includeThoughts: true; retains env fallback for vertex thinkingBudget.

Sequence Diagram(s)

sequenceDiagram
  participant ChatUtil as Chat Utility
  participant Env as Environment
  participant ProviderSDK as Google Provider SDK

  ChatUtil->>ChatUtil: Read model config (thinkingLevel/thinkingBudget)
  ChatUtil->>Env: If google-vertex and thinkingBudget missing -> read GOOGLE_VERTEX_THINKING_BUDGET_TOKENS
  ChatUtil->>ProviderSDK: Build providerOptions { includeThoughts: true, thinkingLevel?, thinkingBudget? }
  ProviderSDK->>ProviderSDK: Initialize request with thinkingConfig
  ProviderSDK->>ProviderSDK: Execute API call

Loading

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• feat(ask_sb): Add headers param to config to allow users to specify custom headers #449: Modifies language model provider schemas/types including GoogleGenerativeAI/GoogleVertex interfaces.
• feat(web): add MCP HTTP endpoint with Streamable HTTP transport #976: Changes the chat utility (packages/web/src/features/chat/utils.server.ts) and provider option handling; overlaps runtime config edits.
• feat(worker,web): Support google secrets as a token type #594: Extends Google provider schema/types (related changes around provider configuration and secrets).

Suggested reviewers

• msukkari
• Ferexx

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'feat: add thinkingLevel and thinkingBudget config for Gemini models' directly and clearly summarizes the main feature addition in the changeset—adding two new configuration properties for Gemini models.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch brendan/gemini-thinking-config-SOU-893

---

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 and usage tips.}