TabbyML: open-source GitHub Copilot alternative with self-hosted codebase RAG
In one sentence TabbyML reaches production maturity with FIM (fill-in-the-middle) completion, local repository RAG indexing, VS Code and JetBrains plugins, and Docker deployment — the first open-source Copilot alternative with awareness of your own codebase.
GitHub Copilot is extraordinarily useful, but it raises an uncomfortable question: do you really want all your company's code — including proprietary parts, critical business logic, credentials accidentally left in comments — sent to Microsoft's servers? TabbyML offers the same AI completion experience, but installed entirely on your own server, sending nothing outside.
The most interesting feature is "codebase context": TabbyML indexes your repository's code and uses that knowledge to suggest completions that adapt to your patterns, your naming conventions, your existing functions. It's not just a generic model that knows Python — it's an assistant that specifically knows your code.
Installation is done with Docker, and the VS Code and JetBrains plugins make the experience identical to Copilot from the user's perspective: you write code, and completion arrives automatically as gray ghost text to accept with Tab. The privacy of proprietary code is 100% guaranteed.
Companies
TabbyML
Tools
—
Tags
Sources