Open Repo Chat

An offline AI-powered code assistant for VS Code. Chat with your codebase using local LLMs via Ollama - no cloud, no API keys, complete privacy.

Features

• Offline and Private - All processing happens locally on your machine
• Chat with Your Code - Ask questions about your codebase and get contextual answers
• Smart Code Search - Uses RAG (Retrieval Augmented Generation) to find relevant code
• Code Generation - Generate code with one-click apply to files
• Conversation Memory - Maintains context across multiple messages
• Fast Local Inference - Powered by Ollama with GPU acceleration

Requirements

• Ollama must be installed and running
• Recommended: 8GB+ RAM for smaller models, 16GB+ for larger models

Quick Start

1. Install Ollama
2. Install this extension
3. Open the Open Repo Chat sidebar (chat bubble icon)
4. Go to Setup tab and verify Ollama is running
5. Click Index Codebase in the Chat tab
6. Start chatting with your code!

Settings

• openRepoChat.chatModel - Ollama model for chat (default: qwen2.5-coder:7b)
• openRepoChat.embeddingModel - Model for embeddings (default: nomic-embed-text)
• openRepoChat.ollamaUrl - Ollama server URL
• openRepoChat.contextChunks - Number of code chunks to include as context
• openRepoChat.temperature - Response creativity (0-1)
• openRepoChat.maxTokens - Maximum response length

Privacy

All data stays on your machine. No telemetry, no API keys required, works completely offline.

Release Notes

1.0.0

Initial release with RAG-powered code chat, conversation memory, and one-click code apply.

License

MIT