Description
Vector-only knowledge bases stumble on cross-document questions and miss what's in charts. NexusRAG is a hybrid RAG system combining vector search, a LightRAG knowledge graph and cross-encoder reranking.
It parses documents with Docling, captions images and tables, cites inline and runs on Gemini or local Ollama.
Hybrid:Vectors plus graph.
Reranking:Cross-encoder precision.
Visuals:Images and tables captioned.
Citations:Sources inline.
It parses documents with Docling, captions images and tables, cites inline and runs on Gemini or local Ollama.
Features
Hybrid:Vectors plus graph.
Reranking:Cross-encoder precision.
Visuals:Images and tables captioned.
Citations:Sources inline.
