DeepRepo

소개

DeepRepo is a production-grade Python library crafted for implementing Retrieval Augmented Generation (RAG) directly on local code repositories. It distinguishes itself with a lightweight, raw Python design, avoiding the overhead of heavy frameworks like LangChain or LlamaIndex and eliminating the need for external vector databases. Offering extensive flexibility, DeepRepo supports a wide range of AI providers including Ollama (for local use), HuggingFace, OpenAI, Anthropic, and Gemini, catering to diverse needs from privacy-focused local development to high-performance production deployments. It also features a RESTful API, Docker readiness, and seamless integration as an MCP server for AI assistants like Cursor and Claude Desktop, empowering developers with full control over their RAG pipelines.

주요 기능

• Raw Python implementation without heavy frameworks or external vector databases
• 0 GitHub stars
• Support for multiple AI providers: Ollama, HuggingFace, OpenAI, Anthropic, and Gemini
• MCP (Model Context Protocol) server for integration with AI assistants like Cursor
• NumPy-powered vector store with cosine similarity and JSON persistence
• RESTful API for easy integration and full Docker containerization

사용 사례

• Building RAG systems for private or offline codebases with local LLMs (e.g., Ollama)
• Integrating codebase knowledge into AI assistants like Cursor or Claude Desktop via MCP server
• Developing flexible and scalable RAG solutions by mixing and matching various cloud and local AI providers