Your LLM clients don't know they're talking to Gemini

What it does This project is a self-hosted proxy that sits between your applications and a growing list of LLM backends. It exposes standard OpenAI and Anthropic API shapes so existing clients—Claude Code, Cursor, SillyTavern, Continue—can talk to Gemini, OpenRouter, or other providers without code changes. Under the hood it also ships as a standalone Python library for managing credential rotation and failover in your own async applications.

The interesting bit The clever part is the dual personality: it is both a drop-in network proxy and a reusable resilience library you can import into your own code. By routing requests with a provider/model_name convention, you can point an Anthropic SDK client at a Gemini model and the translation happens transparently.

Key highlights

• OpenAI- and Anthropic-compatible endpoints for any LiteLLM-supported provider
• Automatic key rotation, rate-limit backoff, and intelligent failover between credentials
• Includes exclusive provider support like Gemini CLI via OAuth
• Interactive TUI for configuration, credential management, and usage stats
• Deployable as a standalone FastAPI service or imported as the rotator_library Python package

Caveats

• README still contains placeholder <!-- TODO --> tags where screenshots should be
• Usage and quota tracking is labeled Alpha
• OAuth providers require local authentication before credentials can be mounted into Docker

Verdict Worth a look if you manage keys across multiple providers or want to route Claude Code to Gemini backends. Skip it if you only ever call a single API from one script.