RAG with ChromaDB

About

This tool functions as a robust Model Context Protocol (MCP) server designed to empower Large Language Models (LLMs) with Retrieval-Augmented Generation (RAG) capabilities. It supports ingesting and processing documents across a wide array of formats, including PDF, Word, Excel, HTML, and more. Utilizing LangChain for its core RAG pipeline and ChromaDB for persistent vector storage, it offers flexible embedding options with either OpenAI for high performance or Ollama for local, free processing. This enables semantic search and retrieval of relevant information from custom datasets, significantly improving the contextual understanding of LLMs.

Key Features

• Multi-format document ingestion with automatic format detection
• Concurrent document processing and batch database writes for efficiency
• Dual embedding support: OpenAI (recommended) or Ollama (free, local)
• Support for URLs, single files, and entire directories as ingestion sources
• Persistent vector storage using ChromaDB
• 0 GitHub stars

Use Cases

• Augmenting LLMs (e.g., Claude Desktop) with custom, up-to-date information from diverse internal documents.
• Building knowledge bases that allow semantic search across an organization's varied document archives.
• Creating an easily deployable RAG backend for applications requiring context from multiple file types.