io.github.HaseebKhalid1507/velociragsearch

velocirag

Verified · yesterday

Lightning-fast RAG for AI agents. 4-layer fusion, ONNX Runtime, sub-200ms search.

Install

claude mcp add velocirag -- uvx velocirag

Our take

VelociRAG is a local RAG engine with MCP tools for indexing Markdown and searching with vector, BM25, graph, metadata, and reranking layers. The source distribution includes real tests and implementation rather than a wrapper stub, and it avoids PyTorch by using ONNX. First use downloads Hugging Face model files into a local cache, so it is not fully offline until warmed.

reviewed by hand · 2026-07-05

Something wrong or dead here? Report it

Verification record

last verified: yesterday
github stars: 10
last commit: 3 mo ago
archived: no
license: MIT
in registry since: 2026-04-05

github.com/HaseebKhalid1507/VelociRAG