Problem

Building emdash requires Python for node-gyp to compile native modules (sqlite3, node-pty, keytar). The current setup has friction points:

1. Inconsistent Python detection: Users with pyenv or multiple Python installations hit errors like "Python is not set from command line or npm configuration"
2. Manual configuration: Developers must manually set PYTHON=/usr/bin/python3 or configure npm globally
3. CI maintenance: Each CI job must install and configure Python separately (see lines 46-54 in release.yml)
4. Onboarding friction: New contributors on different systems (Windows/Linux/macOS) have unpredictable Python setup experiences

Proposed Solution: Use uv

uv is a Python package and project manager written in Rust. It solves the Python environment problem elegantly:

• Reads .python-version or pyproject.toml to pin Python version
• Auto-installs the specified Python version if missing
• Overrides system Python for the project, eliminating conflicts
• Works identically across Windows, Linux, macOS
• 10-100x faster than pip for CI builds

Implementation

1. Add uv to the bootstrap process:

   # For Linux/macOS
   curl -LsSf https://astral.sh/uv/install.sh | sh
   
   # For Windows (Powershell)
   powershell -ExecutionPolicy BypassUser -c "irm https://astral.sh/uv/install.ps1 | iex"

2. Create .python-version:

   3.11.0
   

3. Update release.yml:

   - name: Install uv
     uses: astral-sh/setup-uv@v2
     with:
       version: "0.4.x"
   
   - name: Set up Python
     run: uv python install
   
   - name: Install dependencies
     run: uv pip install -r requirements.txt

4. Create requirements.txt (minimal, just for node-gyp):

   setuptools>=67.0
   wheel
   

5. Update local development docs to mention uv as the recommended Python setup.

Benefits

• Single source of truth for Python version (.python-version)
• Eliminates npm config set python workarounds
• CI becomes faster and more reliable
• New contributors have a consistent experience
• Aligns with industry trends (uv is adopted by major projects: Ruff, Astral tools, etc.)

Industry Adoption

Major projects have adopted uv for Python management:

• Ruff (Python linter)
• Anthropic's various Python projects
• Increasing adoption in JavaScript/Electron projects that need Python build tools

Backwards Compatibility

This is additive and transparent:

• Developers who already have Python set up won't see changes
• Developers without Python get a clear setup path
• CI becomes more deterministic

Questions for Discussion

1. Should we pin Python 3.11 (current CI version) or allow flexibility?
2. Do we want to manage additional Python dependencies beyond node-gyp, or keep it minimal?
3. Should this be documented in CONTRIBUTING.md alongside other setup instructions?