Imported from turquoisehealth/pricepoints (AGENTS.md). Install upstream with npx skills add turquoisehealth/pricepoints. Copyright stays with the author.

Price Points Research Monorepo

What This Is

Price Points creates quantitative, public research on healthcare pricing using Turquoise Health's healthcare price transparency data. This research helps illuminate actual healthcare costs, pricing variation, and market dynamics (e.g., 340B drug pricing, delivery costs, regional disparities).

Monorepo structure:

• projects/ - Time-stamped research projects (e.g., 2025_04_delivery_costs) following the template at 0000_00_proj_template/
• analyses/ - One-off exploratory analyses
• tq/ - Shared Python package for database access and utilities

Standard project layout:

projects/YYYY_MM_project_name/
├── pyproject.toml         # Dependencies
├── analysis.qmd           # Main Quarto notebook
├── ingest.py              # Data ingestion
├── queries/               # SQL files
└── data/                  # Input/output data

Tech Stack

• Python - Version in .python-version (managed with uv, never pip/poetry)
• Data - Polars ONLY (never pandas or numpy)
• Database - Trino via tq package connectors
• Analysis - Quarto (.qmd) for R-based visualization
• SQL - Trino dialect (uppercase keywords, explicit aliases)

Critical Rules

1. Polars only - Do not import or suggest pandas/numpy for data manipulation
2. Use uv - For all dependency management (uv sync, uv add, etc.)
3. Minimal file creation - Only create files when absolutely necessary
4. No documentation - Never proactively create README or .md files
5. Use Claude plugins - When working with Python, invoke the relevant /astral: for uv, ty, and ruff to ensure best practices are followed
6. Pre-commit handles linting - Don't worry about code style, it's automated

Development Workflow

1. Navigate to project directory
2. Run uv sync to install dependencies
3. Write SQL queries in queries/ folder
4. Implement data ingestion in ingest.py (uses tq package for Trino)
5. Analyze in analysis.qmd (Quarto notebook)
6. Let pre-commit hooks handle formatting on commit

Common Patterns

1. Database access:

   from tq.db import get_trino_connection
   conn = get_trino_connection()

2. Starting a new project: Copy 0000_00_proj_template/ to projects/YYYY_MM_name/

3. Testing: Run pytest in tq/ directory when modifying shared utilities

4. Exploration: Use the trino skill to explore available data