asqrzk/copilot-openrouter-to-ollama-proxy

This is a small proxy server that lets GitHub Copilot talk to AI models from OpenRouter instead of Microsoft's backend. OpenRouter is a service that provides access to many different AI models through a single API. Normally Copilot only works with its own servers, but by running this proxy on your computer, Copilot's requests get redirected through OpenRouter, which means you can use models like GPT-4, Claude, or others that OpenRouter supports. The proxy is a fork of an earlier project that did something similar for Ollama (a tool for running AI models locally). This version adds support for GitHub Copilot by exposing the specific API endpoint that Copilot expects. Copilot uses a format called OpenAI-compatible chat completions, so the proxy translates responses from OpenRouter into exactly that format, including proper streaming so answers appear word by word as they are generated. Setup involves setting your OpenRouter API key as an environment variable and starting the proxy with Docker. Once running, the proxy listens on the same port that Ollama normally uses, so any tool configured to point at a local Ollama server will work without further changes. The project also strips out non-standard fields that OpenRouter sometimes adds to responses, since Copilot is strict about only accepting spec-compliant payloads. If a model returns a reasoning section (from thinking-style models), the proxy appends that content to the main message so Copilot can display it. The README is brief and links to the original repository for more detailed setup instructions. The license is not mentioned.