PDF Retrieval

About

This tool functions as a Model Context Protocol (MCP) server, offering a completely free and efficient way to extract relevant information from PDF documents. It employs a sophisticated hybrid search mechanism, combining BM25 keyword matching with semantic vector search, to ensure highly accurate retrieval of document chunks. Designed for seamless integration with AI agents, it processes PDFs automatically and returns raw text chunks, empowering agents to leverage document context without incurring API costs for embeddings or LLM answer generation.

Key Features

• Free embeddings using ChromaDB's default sentence-transformers (no API costs)
• 0 GitHub stars
• Automatic parsing and indexing of PDF files using Docling
• Exposes `retrieve_pdf_chunks` tool via FastMCP for seamless agent integration
• Pure retrieval mode returns raw document chunks for agent processing
• Hybrid retrieval combining BM25 and vector search for accuracy

Use Cases

• Building knowledge retrieval systems for internal documentation or research papers
• Empowering AI agents to answer user questions based on private PDF document content
• Providing contextual information from documents to large language models for informed responses