{"repo":"arbs-io/github-copilot-llm-gateway","free":true,"listed":false,"github":"https://github.com/arbs-io/github-copilot-llm-gateway","clone":"git clone https://github.com/arbs-io/github-copilot-llm-gateway.git","description":"GitHub Copilot LLM Gateway is a companion extension for GitHub Copilot that adds support for self-hosted open-source models. It seamlessly integrates with the Copilot chat experience, allowing you to use models like Qwen, Llama, and Mistral alongside or instead of the default Copilot models.","language":"TypeScript","stars":67,"topics":[],"license":"MIT","category":"ai-agents","readme_excerpt":"GitHub Copilot LLM Gateway A robustness layer for running self-hosted open-source models inside GitHub Copilot Chat — built for the models and servers that don't quite behave. Do I need this, or is native BYOK enough? Since VS Code 1.122 , VS Code ships a built-in BYOK \"Custom Endpoint\" provider (Generally Available) that connects any OpenAI-compatible server — vLLM, Ollama, llama.cpp, LM Studio, LocalAI — directly to Copilot chat, agent mode, tools, and MCP, with no extension and no GitHub sign-in required . For most setups that's the simplest path, and you should start there: run Chat: Manage Language Models from the Command Palette and add a Custom Endpoint. This extension is for the harder cases native BYOK doesn't handle. Native BYOK trusts your endpoint as-is and does no quirk-smoothing — its own docs note that tool-call reliability \"depends on your server's tool-call parser.\" When you're stuck with a specific small or quantized model, or a server you can't reconfigure, that's where this extension earns its place: Use native BYOK (built in) when… Use this extension when… -------------------------------------------------------------------------- ---------------------------------------------------------------------------------- Your model + server are well-behaved Tool calls fail with malformed / truncated JSON You can pick the model and configure the server (e.g. --tool-call-parser ) You're locked to a specific small / quantized model that emits sloppy tool calls You wan","default_branch":null,"files":null,"tree":[],"storefront":"/r/arbs-io","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/arbs-io/github-copilot-llm-gateway/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}