LeDat98/NexusRAG

Hybrid RAG system combining vector search, knowledge graph (LightRAG), and cross-encoder reranking — with Docling document parsing, visual intelligence (image/table captioning), agentic streaming chat, and inline citations. Powered by Gemini or local Ollama models.

45
/ 100
Emerging

The system employs a two-tier embedding architecture—BAAI/bge-m3 (1024-dim) for fast vector retrieval and a configurable second model (Gemini 3072-dim, Ollama, or sentence-transformers) exclusively for knowledge graph entity extraction—optimizing each stage for its computational requirements. Docling or Marker parsers preserve document structure (headings, page boundaries, tables) and auto-caption images/tables for semantic searchability before chunking, while LightRAG extracts entity relationships for multi-hop traversal independent of vector similarity. The frontend (React 19) integrates a document viewer with interactive knowledge graph visualization, allowing users to navigate citations back to source pages and explore entity connections discovered during retrieval.

179 stars.

No License No Package No Dependents
Maintenance 13 / 25
Adoption 10 / 25
Maturity 1 / 25
Community 21 / 25

How are scores calculated?

Stars

179

Forks

43

Language

Python

License

Last pushed

Mar 17, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/rag/LeDat98/NexusRAG"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.