What is twinny?

Twinny is a free extension for Visual Studio Code that adds AI-powered code completion and a chat panel to the editor. It works with locally running AI models as well as cloud-based API providers, so you can use it entirely offline or connect it to a service like OpenAI, Anthropic, Groq, Mistral, or others. The two main features are code completion and a chat sidebar. The completion feature watches what you are typing and suggests how to finish the current line or block of code, similar to how GitHub Copilot works. The chat sidebar lets you highlight a piece of code, ask a question about it, and get an explanation, a refactored version, or a test written for it, all without leaving the editor. Chat history is saved between sessions. Twinny also includes some smaller utilities: it can generate a suggested commit message based on your staged changes, open a side-by-side diff view, and create a new document from a code block the AI produces. You can accept a solution directly into your file or copy it to the clipboard. For local use, Twinny connects by default to any OpenAI-compatible API running on your machine, which includes tools like Ollama. You point it at the address and port where your local model is listening, choose the model name, and it works. If you prefer a cloud provider, you switch the provider setting and enter an API key. The README lists nine supported providers and notes that custom API endpoints can be configured. A feature called workspace embeddings lets Twinny index your project files so the AI has context about the broader codebase when answering questions, rather than only seeing what is in the current file. The README also mentions a peer-to-peer network called Symmetry for sharing AI inference resources, though the description of that feature is brief.