CLINI-Q — Intelligent SOP Navigator (Streamlit MVP)

This is a minimal, deployable Streamlit app that demonstrates the CLINI‑Q concept: an AI‑assisted SOP navigator that turns role‑specific questions into actionable procedural guidance with citations to site SOPs.

MVP Guardrails: No medical advice. No PHI/PII. Not an IRB/regulatory submission tool. Verify locally with PI & SOPs.

Features

• Role selection (CRC, RN, Admin, Trainee) and scenario entry
• TF‑IDF retrieval over local SOP files (/data/sops/*.txt|*.pdf)
• Optional OpenAI drafting of step‑by‑step guidance with citations (falls back to offline mode)
• Clear compliance reminders and disclaimers

Quickstart (Local)

python -m venv .venv && source .venv/bin/activate  # or .venv\Scripts\activate on Windows
pip install -r requirements.txt
# (optional) export OPENAI_API_KEY=sk-...

streamlit run app.py

Place your SOP files under data/sops/. You can use python ingest.py <path/to/file.pdf> to copy files into the repo.

Deploy to Streamlit Community Cloud

1. Push this folder to a GitHub repo (e.g., cliniq-streamlit).
2. In Streamlit Cloud, create a new app:
   • Repository: yourname/cliniq-streamlit
   • Branch: main
   • Main file path: app.py
3. In Settings → Secrets, add:

   OPENAI_API_KEY = "sk-..."

4. Click Deploy.

Configuration

• OPENAI_MODEL env var (optional) chooses the model (default gpt-4o-mini).
• Evidence snippets shown per query are controlled by the sidebar slider.

Project Structure

.
├── app.py
├── ingest.py
├── requirements.txt
├── .streamlit/
│   └── secrets.toml (example)
└── data/
    └── sops/
        └── sample_sop.txt

Notes

• PDF text extraction uses pypdf. Ensure your PDFs are text‑based, not pure scans.
• Retrieval uses scikit‑learn TF‑IDF; replace with vector search later if needed.
• The app outputs procedural guidance and citations to file names; always verify with your SOPs.
• Licensed MIT.