README.md

metadata

title: Clinical Research Basics
emoji: 🧪
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 5.49.0
app_file: app.py
pinned: false

🧪 Clinical Research Chatbot

A lightweight, fully open-source chatbot for clinical research professionals.
Runs entirely on Hugging Face — no OpenAI dependency.

---

✅ Current Features

💬 Chatbot Interface

• Gradio UI with chatbot + Admin Tools tab.
• Query pipeline: FAQ → Glossary → Knowledge Base → APIs (PubMed → FDA → ClinicalTrials.gov).
• Answers are clearly labeled by source.

🔍 Knowledge Base (Docs + URLs)

• Supports ingestion of: PDF, DOCX, TXT, XLSX, JSON, HTML.
• Auto-ingests from:
  • /data/public_docs/
  • /data/urls.txt
• Smart chunking optimized for glossary terms + long text.

📦 Vector Search

• FAISS + all-MiniLM-L6-v2 embeddings.
• Persistent storage:
  • /persistent/faiss.index
  • /persistent/faiss.index.meta.json
• Index survives restarts and can be exported/imported as .zip.

🌐 API Integrations

• PubMed
• FDA Drug Labels
• ClinicalTrials.gov

🧠 Query Handling

• Glossary-aware normalization
  (e.g., eCRF, e-CRF, electronic case report form → same match)
• Glossary priority: if glossary hit exists → always returned first.
• Answer flow: FAQ → Glossary → KB → APIs.
• Clear section labels, citations, and confidence notes.

📜 Logging

All queries, answers, and sources saved in: