YDRAG FAQs

Question 1

Can YDRAG integrate with AI assistants or other applications?

Accepted Answer

Yes, YDRAG features an MCP (Model Context Protocol) server. It offers configurable transports (stdio, SSE, Streamable HTTP), making it straightforward to integrate with various AI assistants like Claude, Amp, or other custom tools.

Question 2

Which database does YDRAG use for vector storage and search?

Accepted Answer

YDRAG utilizes DuckDB as its vector database. It persistently stores documents and their embeddings and performs rapid vector similarity searches using DuckDB's native `array_cosine_similarity` function.

Question 3

How does YDRAG handle text embedding generation?

Accepted Answer

YDRAG leverages the YZMA library (a Go binding for llama.cpp) to generate text embeddings locally. This allows users to utilize any GGUF embedding model, ensuring data privacy and control over the embedding process.

Question 4

What are the core prerequisites for running YDRAG?

Accepted Answer

To run YDRAG, you will need Go 1.24+ (with CGo enabled), a C compiler (gcc or clang), the llama.cpp library, and a GGUF embedding model. YDRAG also supports PDF text extraction for easy document ingestion.

Question 5

What is YDRAG and its primary purpose?

Accepted Answer

YDRAG is a Go-based Retrieval-Augmented Generation (RAG) system designed to build local knowledge bases. It performs local text embedding generation via YZMA/llama.cpp and uses DuckDB for efficient vector storage and similarity search, enabling robust AI assistant integration.

YDRAG

YDRAG

主要功能

使用案例

主要功能

使用案例