YDRAG

Acerca de

YDRAG is a versatile Retrieval-Augmented Generation (RAG) system built in Go, designed to enhance AI assistant capabilities with custom knowledge bases. It leverages YZMA for efficient local embedding generation using GGUF models, employs DuckDB as a fast, in-process vector database for similarity search, and integrates seamlessly with AI assistants via the Model Context Protocol (MCP). This system empowers users to ingest, query, and manage their documents, providing relevant context to large language models for more accurate and informed responses, all while supporting flexible deployment options through various server transports.

Características Principales

• Perform vector similarity search with DuckDB's native array_cosine_similarity for efficient retrieval
• Persistently store documents and their embeddings for a reliable knowledge base
• Extract text from PDF documents to facilitate easy ingestion
• 0 GitHub stars
• Generate local text embeddings using any GGUF embedding model via YZMA/llama.cpp
• Provide an MCP server with configurable transports (stdio, SSE, Streamable HTTP) for AI assistant integration

Casos de Uso

• Augmenting AI assistants with custom, up-to-date knowledge bases for more accurate responses
• Building local and privacy-focused RAG systems without external cloud dependencies
• Managing and semantically querying personal or project documentation to quickly find relevant information