What this PR does

Before this PR:
Gemini 3 models (e.g., gemini-3-flash-preview) did not include the reasoning_effort parameter in HTTP requests when deep thinking mode was enabled through OpenAI Compatible providers. The parameter was silently dropped.

After this PR:
Gemini 3 models correctly send the reasoning_effort parameter when using OpenAI Compatible providers.

Fixes #13221

Why we need it and why it was done in this way

The AI SDK's @ai-sdk/openai-compatible provider schema only recognizes camelCase reasoningEffort, not snake_case reasoning_effort. When snake_case was passed via provider options, it failed schema validation. The SDK then explicitly mapped reasoning_effort: compatibleOptions.reasoningEffort in the request body, which resolved to undefined (since the schema rejected the snake_case key), overriding any spread value and silently dropping the parameter.

The fix changes the return value from { reasoning_effort: reasoningEffort } to { reasoningEffort } (camelCase) so the AI SDK schema recognizes it and properly maps it to reasoning_effort in the HTTP request body.

The following tradeoffs were made: N/A

The following alternatives were considered: Patching the AI SDK to accept snake_case — rejected because it's a third-party dependency and the camelCase format is the documented API.

Breaking changes

None

Special notes for your reviewer

The existing comment in src/renderer/src/types/sdk.ts (line 82) already warns about this exact behavior: // WARN: This field will be overwrite to undefined by aisdk if the provider is openai-compatible. Use reasoningEffort instead.

Checklist

• PR: The PR description is expressive enough and will help future contributors
• Code: Write code that humans can understand and Keep it simple
• Refactor: You have left the code cleaner than you found it (Boy Scout Rule)
• Upgrade: Impact of this change on upgrade flows was considered and addressed if required
• Documentation: Not required — bug fix with no behavior change from the user's perspective
• Self-review: I have reviewed my own code before requesting review from others

Release note

Fix Gemini 3 models not sending reasoning_effort parameter when deep thinking mode is enabled through OpenAI Compatible providers.