Knowledge RAG

A local knowledge retrieval system that combines semantic search with keyword-based routing for document analysis across security, development, and general knowledge domains. Uses ChromaDB for vector storage, Ollama for local embeddings, and implements hybrid search through Reciprocal Rank Fusion combining BM25 keyword matching with semantic similarity. Supports multiple document formats (PDF, Markdown, code files) with intelligent chunking and category detection, featuring specialized keyword routing for cybersecurity topics like red team tactics, blue team detection, CTF challenges, and WAF bypass techniques.

Source

Repository: https://github.com/lyonzin/knowledge-rag