← All projects
Machine learning & AI · Python
R

RAG Document QA

Ask questions about your documents and get an answer with citations. Retrieval-augmented generation without the generation bill: chunking, hybrid BM25 + TF-IDF retrieval, and an extractive answerer that can only quote what is in your files — all in your browser.

GitHub

01 Ask

documents → chunk (by heading, with overlap) → index (BM25 + TF-IDF) → top-k chunks → extractive answer + citations
Ask a question to see an extractive answer built only from sentences in the corpus.

    02 Indexed corpus

    Three documents from the sample corpus: the employee handbook, expenses & travel, and engineering onboarding. Chunking splits on markdown headings, so a chunk never crosses a topic boundary — or paste your own below and ask about it instead.