Local RAG Backend
Enables local RAG systems by registering diverse document formats for advanced search capabilities through an MCP server.
About
This Docker Compose-based backend provides a comprehensive RAG (Retrieval-Augmented Generation) system for local environments. It supports the ingestion of documents in 28 different file formats using `unstructured`, creating a rich knowledge base. Leveraging `graphiti` and a `Neo4j` database, it offers sophisticated search capabilities combining vector, graph, and full-text search, with results intelligently reranked based on relevance. A unique feature is its ability to track changes in conceptual relationships within documents over time through `graphiti`'s episode functionality, making it ideal for evolving knowledge domains accessible via an MCP server.
Key Features
- Provides Model Context Protocol (MCP) compatible search functionality using `graphiti MCP Server`.
- Runs entirely on Docker Compose for easy local deployment and management.
- Supports 28 document file formats (PDF, Office, text, image, etc.) via `unstructured` for broad compatibility.
- 0 GitHub stars
- Offers advanced RAG with combined vector, graph, and full-text search, plus relationship-based reranking.
- Tracks temporal changes in document concept relationships using `graphiti`'s episode feature.
Use Cases
- Establishing a private, local knowledge base for Retrieval-Augmented Generation applications.
- Analyzing the evolution of concepts and relationships within document collections over time.
- Providing AI agents and MCP Clients with sophisticated search capabilities across diverse internal documents.