Modular RAG MCP Server

Example Use Cases

• 1

  A modular and resilient RAG system with hybrid retrieval and reranking.

• 2

  Multimodal document understanding including PDF tables and image captioning.

• 3

  Integration with MCP-compatible AI assistants via Stdio or HTTP transport.

Description

This MCP server provides a complete RAG pipeline including document ingestion, hybrid search combining BM25 and dense embeddings, reranking, and multimodal capabilities such as PDF table extraction and image captioning. It supports zero-code component switching via configuration, four-layer API resilience for robust operation, and dual transport modes (Stdio and HTTP) for flexible deployment. The system offers full observability with JSONL tracing and a Streamlit dashboard, making it suitable for integration with AI assistants like GitHub Copilot and Claude Desktop.