pycharm-runnable-python-deliverable

name	pycharm-runnable-python-deliverable
description	Use when the user wants a Windows/PyCharm-ready Python deliverable such as a single runnable .py file, small project folder, training/inference script, smoke test, or packaged demo; emphasizes inspecting real assets, dependency compatibility, non-ASCII path safety, and runnable verification.

PyCharm Runnable Python Deliverable

Core Workflow

1. Inspect the real folder before coding:
   • datasets,
   • weights,
   • split files,
   • configs,
   • existing modules,
   • expected output format.
2. Decide whether the user needs training, inference, UI integration, or a one-file utility.
3. Check interpreter and dependency compatibility early, especially when the user requests exact versions.
4. Prefer a directly runnable .py file or compact project folder over abstract snippets.
5. Make paths robust on Windows:
   • use Path(__file__).resolve().parent,
   • support Chinese/non-ASCII paths when images or datasets are involved,
   • avoid hardcoded absolute paths unless the user explicitly wants them.
6. Add command-line options only when they help the user test or scale the run.
7. Include a tiny smoke-test path or sample limit so the script can be verified quickly.
8. Run the smoke test and report the command and output artifacts.

Decision Rules

• If existing weights are present and the user says not to retrain, build inference or UI around them.
• If a dependency version requires a newer Python, explain the interpreter floor before trying many incompatible installs.
• For computer vision scripts on Windows, prefer byte-based OpenCV IO when path encoding may be a risk.
• Keep changes line-level when the existing file is close to correct.
• For PyCharm delivery, include stable working-directory behavior and clear run instructions.

Verification

• The final script or project runs from the intended working directory.
• Imports succeed in the selected interpreter.
• Input discovery works against the real local assets or included samples.
• A smoke run completes with small sample limits when full training or full inference would be slow.
• Expected artifacts such as checkpoints, CSV logs, preview images, result folders, or console summaries are created.
• The final answer tells the user exactly which file to open or run in PyCharm.