coldpy 0.1.0

Analyze Python startup import performance and memory usage

ColdPy

ColdPy profiles Python startup import cost by measuring import time and memory per module. It is built for serverless, microservices, and large Python apps where cold start matters.

What it measures

• Import duration (milliseconds)
• Import memory peak (MB) using tracemalloc
• Per-module status (success/error)
• Heavy module hints for lazy loading opportunities

Safety caveat

ColdPy measures real imports, so target module import code is executed in isolated subprocesses. This prevents state leakage between modules, but import side effects in target code may still occur.

Install

uv sync

Usage

coldpy scan
coldpy scan ./myproject
coldpy scan ./myproject --json report.json
coldpy scan ./myproject --exclude "migrations/**" --exclude "scripts/**"
coldpy top 10
coldpy top 20 --sort memory --threshold-mb 20

Commands

• coldpy scan [PATH=. ] [--json OUTPUT_JSON] [--threshold-ms N] [--threshold-mb N] [--no-cache]
• coldpy scan PATH [--python PYTHON] [--env-file ENV_FILE] [--no-project-env] [--exclude PATTERN]
• coldpy top [N=10] [--sort time|memory] [--threshold-ms N] [--threshold-mb N]

coldpy top reads ./.coldpy/cache.json and fails if no cache exists.

For scan, ColdPy auto-detects project virtualenv Python in .venv, venv, or env and auto-loads environment variables from .env/.env.local when present. Use --python and --env-file only when you need to override auto-detection. ColdPy also excludes common migration paths by default (alembic/**, migrations/**).

JSON schema (v1)

{
  "schema_version": "1.0",
  "generated_at": "2026-02-20T10:00:00+00:00",
  "project_root": "/path/to/project",
  "settings": {
    "threshold_ms": 100,
    "threshold_mb": 50,
    "exclusions": ["tests", "venv"]
  },
  "summary": {
    "total_modules": 3,
    "scanned_modules": 2,
    "failed_modules": 1
  },
  "modules": [
    {
      "name": "pkg.fast",
      "file": "/path/to/project/pkg/fast.py",
      "import_time_ms": 1.234,
      "memory_mb": 0.123,
      "status": "ok",
      "error": null,
      "notes": []
    }
  ]
}

Cache behavior

• Cache path: ./.coldpy/cache.json
• Written by scan by default
• Disable with --no-cache
• Read by top

Troubleshooting

• Cache not found: run coldpy scan <path> first.
• No Python modules found: verify path and exclusions.
• Import failures: check module side effects and importability from project root.

PyPI release setup

GitHub workflows are included in /Users/denis/Documents/ColdPy/.github/workflows:

• ci.yml: test matrix on Python 3.10/3.11/3.12.
• package.yml: build and validate sdist + wheel.
• publish.yml: publish on GitHub Release (PyPI) or manual dispatch (TestPyPI/PyPI).

Recommended publishing model is Trusted Publisher (OIDC):

1. Create a PyPI project for coldpy (and optionally a TestPyPI project).
2. In PyPI project settings, add your GitHub repo as a trusted publisher.
3. Create a GitHub Release to publish to PyPI.
4. Use workflow dispatch with target=testpypi for preflight package checks.

If you prefer API tokens instead of OIDC, set password input in the publish action and store the token in GitHub Secrets.