FOLDER_STRUCTURES.md

Folder Structure (Key Files)

• app/
  • env.py # .env loading and observability setup
  • main.py # FastAPI app with /answers, /answers/stream, and /process-files
  • models.py # Pydantic models for Patient, QuestionSet, Answer, ProcessFiles, etc.
• sample_data/
  • patient_data.json # Synthetic patient records (list)
  • zepbound_question_set.json # PA question set for demo
• tests/
  • test_answers.py # Live integration test (gated by RUN_LIVE_OPENAI)
  • generate_mock_data.py # Generate mock test cases with ground truth
  • eval_pipeline.py # Evaluation pipeline with F1 score metrics
  • mock_test_data.json # Generated mock test data (created by generate_mock_data.py)
  • eval_results.json # Evaluation results (created by eval_pipeline.py)
  • README.md # Documentation for evaluation pipeline
• streamlit_app.py # Streamlit demo UI (Standard + Streaming + File Upload)
• openai_structured_ouput_official_doc.md # Reference patterns for structured outputs
• homen_shum_notes.md # Project notes / goals
• DESIGN_SPECS.md # Architecture and design decisions for this demo
• FOLDER_STRUCTURES.md # This file

Run Instructions (dev)

1. Start API:
   • uvicorn app.main:app --reload
2. Run Streamlit UI:
   • streamlit run streamlit_app.py
3. Optional live test:
   • set RUN_LIVE_OPENAI=1 && pytest -q
4. Generate mock test data:
   • python -m tests.generate_mock_data
5. Run evaluation pipeline:
   • python -m tests.eval_pipeline
   • python -m tests.eval_pipeline --validate # with validation pass

Dependencies

• FastAPI/Starlette (already in pyproject)
• OpenAI >= 1.86.0 (declared in dev deps)
• Streamlit for demo UI
• requests for HTTP calls from Streamlit

If not present, install via your environment's package manager (uv, pip, etc.).