Local RAG Knowledge Assistant

Local RAG Knowledge Assistant — Retrieval-augmented chatbot, fully offline

• Python pipeline: scrapes and cleans web sources (Trafilatura/BeautifulSoup), chunks documents, embeds with a

multilingual e5 model, and stores vectors in ChromaDB.

• Answers run through a local Llama 3.2 model via Ollama — no data leaves the machine — with strict context-only

prompting and inline source citations to keep it from making things up.

• Curated the corpus toward web-security material (write-ups, references) so it doubles as a study aid for the bounty

work. Includes a small benchmark script to compare retrieval settings.