gpt-load

Why we included this project

GPT-Load is a Go-based transparent proxy for applications that call several AI providers at once. It forwards requests to OpenAI, Gemini, or Claude in each provider's native format, so the client code you already have keeps working while all traffic funnels through one stable endpoint. Behind that endpoint, a key pool rotates credentials, blacklists keys that start failing, and reintroduces them once they recover, which spares you the mid-request throttling a single overloaded key can cause. Weighted load balancing spreads calls across upstreams, and the leader-follower mode supports horizontal scaling for high-concurrency production traffic. The Vue admin panel shows real-time stats, health checks, and request logs, and you can hot-reload group configurations without restarting the service.