MI Browse (Cornerstone)

A minimal FastAPI + Jinja application for exploring the Modality Cornerstone pilot dataset. The app loads Cornerstone session data from local JSON files and provides a small UI for browsing, filtering, and inspecting conversations and MI findings.

Quickstart

1. Create and activate a virtual environment:

   python -m venv .venv
   source .venv/bin/activate

2. Install dependencies:

   pip install -r requirements.txt

3. (Optional) Configure environment overrides:

   cp .env.example .env

4. Ensure Cornerstone data is available:
   • data/cornerstone_data/conversation_*.json
5. Start the development server:

   uvicorn app.main:app --reload

The app will be available at http://127.0.0.1:8000/. Navigate to /datasets/cornerstone to browse sessions.

Key Features

• Session view that groups turns into a readable transcript
• Findings integration for MI skills evaluations (when present)
• User-level views for reviewing MI skills across sessions

Cornerstone Data Layout

The app expects the following files under data/cornerstone_data/:

• conversation_*.json - Raw conversation files used for browsing, findings, and progress reports

MI Skills Evaluation (Optional)

You can generate MI findings for Cornerstone conversations using the OpenAI API.

export OPENAI_API_KEY=your_key_here
python scripts/evaluate_mi_skills.py "data/cornerstone_data/conversation_*.json"

To inspect prompts and schemas without calling the API:

./scripts/test_prompts.sh

Progress Reports (Profiles)

After MI findings are generated, you can build per-user progress reports.

# List available users (from evaluated conversations)
python scripts/generate_progress_reports.py --list-users

# Generate a markdown-only report for a specific user
python scripts/generate_progress_reports.py --user "Full Name" --markdown-only

# Generate AI-assisted progress analysis for a specific user
python scripts/generate_progress_reports.py --user "Full Name"

# Generate AI-assisted reports for all users
python scripts/generate_progress_reports.py --all

Project Layout

app/                FastAPI application, templates, static assets
data/cornerstone_data/   Cornerstone JSON files
scripts/            Cornerstone data and evaluation helpers
tests/              Smoke tests for routes

Run the automated checks with:

PYTHONPATH=. pytest

PYTHONPATH is required so the test runner can import the local app package.

Known Limitations

• Cornerstone data is unlabeled; MI findings are model-generated.
• Findings availability depends on which conversations have been evaluated.