Inventoried all ten Nx projects tagged lang:python, their Ruff dependencies, project-local settings, Pixi tasks, and Nx lint targets.
Measured Ruff C901 McCabe complexity for every function in each project’s production source root with an isolated C901-only invocation.
Found 1,710 functions, a median complexity of 2, and 20 functions above 15. The maximum is 30 in vxdata-jobs.
Found that dashboard-api has no Ruff setup or Nx lint target, and that vxdata-api’s include = ["src"] causes its current lint task to scan no Python files.
Made no changes to the monorepo.
Lessons Learned: Pitfalls
The root README inventory is stale: it omits several current Python projects. Nx project.json tags are the reliable inventory.
Running C901 through each project configuration can silently omit vxdata-api; use --isolated for a consistent audit, then separately inspect the actual lint task.
Ruff versions vary significantly between independent Pixi locks, although the C901 results were stable across the installed versions used here.
Lessons Learned: Improvements
Document a workspace-wide Ruff baseline and standard lint scope for Python projects.
Add a check that each lang:python project has a working lint target and that Ruff discovers at least one Python file.
A McCabe ceiling of 15 is a practical target: it affects about 1.2% of current production functions and can be introduced with a short legacy baseline before ratcheting toward 10.
