gemini-cli - 💡(How to fix) Fix Support VS Code Native AI Chat Interface `chatLanguageModels` via 3-Legged Auth

What would you like to be added?

Integrate Gemini with the native VS Code Language Model Chat Provider API. Allow the official Gemini extension to register itself as a native model provider, routing the native chatLanguageModels traffic through the extension’s existing 3-legged OAuth (3LO) infrastructure.

Why is this needed?

The current official Gemini Code Assist extension forces developers into a custom sidebar interface. Many developers prefer the native VS Code Chat interface (Cmd+Ctrl+I), which provides a unified UX across multiple AI engines and native workspace context handlers.

Currently, to use Gemini in the native VS Code chat UI, developers must fall back to using a standard AI Studio API key (2-legged auth). This severely breaks the developer experience by imposing low, non-subsidized rate limits and causing frequent Tokens-Per-Minute (TPM) caps during heavy operations like @workspace indexing.

Additional context

Benefits:

• Best of Both Worlds: Developers get the 100% native VS Code UX, shortcuts, and UI consistency, while retaining the high-limit, subsidized developer quotas.
• Frictionless Workflow: Bypasses the custom sidebar UI without risking constant "Rate Limit Exceeded" errors.
• Higher Adoption: Attracts power-users who refuse to use custom sidebars but want to leverage Gemini's large context windows.
• Consistent Telemetry: Because requests still pass through the extension's authenticated 3LO wrapper, Google maintains telemetry loops (context acceptance, IDE metrics) that are lost when developers use generic API keys.
• Parity with Competitors: Ensures Gemini remains a first-class citizen as other major AI providers lean into native VS Code subsystems (i.e. Claude Code).

Gemini Code Assist Version: 2.81.0