RAG Knowledge Assistant & Care Navigator

RAG Knowledge Assistant & Care Navigator An independent AI project that brings public clinic information into a source-linked conversation and supports structured doctor-directory navigation. What I built • Python and FastAPI APIs with a browser chat interface. • Hybrid BM25 and embedding retrieval, combined with reciprocal-rank fusion. • Directory tools that filter doctors by speciality, branch and published days. • A bounded model/tool loop and validated booking-request drafts. Evaluation The committed corpus has 343 records and 427 chunks. On a curated 30-question retrieval evaluation, hit@1 was 0.83, hit@3 0.93 and MRR 0.889. These measure retrieval, not generated-answer accuracy. The repository reports 117 Python and 13 Node tests passing on 13 October 2026. Demo: https://anamikasingh.vercel.app/bots Source code: https://github.com/anamika-singh-nw/rdbot Scope Independent work, unaffiliated with RxDx. Uses public clinic information; not a medical or clinical service. Booking outputs are drafts and do not create appointments.