cuda_doctor/cli.py

This file is the Python face of the tool. It parses `check` or `doctor`, validates that `auto` only accompanies `doctor`, resolves `cuda-doctor-core`, and then shells out to the native executable without trying to re-implement the logic in Python.

pyproject.toml main.cpp

Reference

Forwarding contractpython

binary = resolve_core_binary()
forwarded = [str(binary), args.command]
if args.mode is not None:
    forwarded.append(args.mode)
if args.json:
    forwarded.append("--json")
completed = subprocess.run(forwarded, check=False)

What it does

Checks for a local `.dist/cuda-doctor-core` first.
Falls back to a globally discoverable `cuda-doctor-core` on PATH.
Parses the public CLI surface with argparse.
Forwards the exit code from the native executable back to the shell.

Caveats

The Python wrapper does not contain fallback diagnosis logic. If the native binary is missing, the wrapper fails immediately. That keeps the responsibilities clean but means packaging and native build steps matter.

LICENSE

src/main.cpp

Related files

Packaging

pyproject.toml

Defines the Python package metadata and registers the `cuda-doctor` console script entry point.

Native entrypoint

src/main.cpp

Owns native argument validation, text and JSON rendering, exit codes, and the interactive handoff from `doctor` to `doctor auto`.