twinnydotdev/twinny

Twinny – AI‑powered VS Code extension

What it is – Twinny is a free, open‑source Visual Studio Code extension that adds AI‑driven coding helpers directly inside the editor. It talks to any OpenAI‑compatible API (local Ollama, OpenAI, Anthropic, Mistral, etc.) and can also run offline against a self‑hosted model.

Key capabilities

• Fill‑in‑the‑middle completion – real‑time suggestions that insert code at the cursor or replace a selected block, supporting both single‑line and multi‑line fills.
• Chat sidebar – a persistent chat window where you can ask the model to explain code, write tests, refactor, generate commit messages, or create new files from code snippets.
• Workspace embeddings – the extension builds vector embeddings of the current project so the model can retrieve relevant context, making suggestions more accurate.
• Symmetry network – optional peer‑to‑peer (P2P) layer that lets multiple users share inference resources, reducing reliance on a single API endpoint.
• Customization – you can point Twinny at any compatible endpoint, pick the model, adjust prompts, and toggle features such as diff view, full‑screen chat, or automatic code‑solution acceptance.

Supported back‑ends – OpenAI, Anthropic, OpenRouter, Deepseek, Cohere, Mistral AI, Perplexity, Groq, and any locally hosted OpenAI‑compatible server (e.g., Ollama).

Installation & docs – Install from the VS Code Marketplace or via the repo, then follow the quick‑start guide at the official docs site.

Community – MIT‑licensed, welcomes contributions via GitHub issues and pull requests, and provides a donation address for supporters.

---

Twinny aims to bring the convenience of modern LLM assistants to everyday coding workflows while keeping the option to run entirely offline.