AGENTS-python.md

AGENTS-python.md

Purpose

Instructions for Codex working on Python machine learning experiments in this repo.

Environment

• Use Python with pyenv, pip, and git.
• Prefer solutions that work on Linux and macOS.

Priorities

• This is experiment code, not production code.
• Prefer code that is simple, short, and straightforward.
• Optimize for fast iteration and easy edits.
• Do not over-engineer robustness, abstractions, or configuration.

Style

• Format code with black using line length 120.
• Sort imports with isort.
• Keep functions small and direct.
• Prefer scripts and plain functions over complex class structures.

Logging

• Add lightweight print() logging for important experiment steps.
• Log enough intermediate results to visually verify behavior.
• Prefer concise logs over noisy logs.

Validation

• Run the smallest useful check for the change.
• If formatting is needed, use black --line-length 120 . and isort ..

Safety

• Never overwrite user changes you did not make.
• Never use destructive git commands unless explicitly requested.