Healthcare Assistant (Local, Ollama)

A local-first chatbot to understand your healthcare data (labs PDFs, medications CSV/XLSX, WHOOP CSVs) using LangChain + Ollama + Chroma.

Features

• Ingests:
  • PDFs in data/labs/
  • Meds CSV/XLSX in data/medications/
  • WHOOP CSVs in data/whoop/
• Builds a vector index (Chroma) from a lightweight text corpus.
• Streamlit chat UI with basic RAG and citations.
• No external API costs (uses Ollama + sentence-transformers).

Quickstart

1. Install Ollama and a model:
   • https://ollama.com/download
   • Recommended chat model: ollama pull qwen3:14b
2. Python env
   • python -m venv .venv && source .venv/bin/activate
   • pip install -r requirements.txt
3. Configure env
   • cp .env.example .env and adjust if needed
4. Ingest data
   • python -m app.ingestion.ingest_labs
   • python -m app.ingestion.ingest_meds
   • python -m app.ingestion.ingest_whoop
5. Build index
   • python -m app.indexing.build_index
6. Run UI
   • streamlit run app/app.py
   • Use the --logger.level=info flag to see more detailed logging

Notes

• This app does not provide medical advice. Use for education and discussion with clinicians.
• If no corpus/index exists, the chat will prompt you to run ingestion+indexing.