Стандарты Python для AI-монорепозитория

learning

python-skills is a Claude Code skill that provides a shared Python best-practices reference for every component of the LlamaFarm monorepo — Server (FastAPI, Celery, Pydantic, structlog), RAG (LlamaIndex, ChromaDB), Universal Runtime (PyTorch, transformers), Config, and Common. The skill is non-user-invocable and operates with Read, Grep, and Glob tools. It bundles six topic files — patterns.md, async.md, typing.md, testing.md, error-handling.md, and security.md — plus a prioritised code-review checklist covering everything from modern type hints and asyncio correctness to Critical-priority security checks such as path traversal prevention and safe deserialization. Code style is enforced via ruff with rule sets E, F, I, B, UP, and SIM targeting Python 3.11, line length 88. Ideal for teams building AI infrastructure who need consistent async patterns, pytest-asyncio testing conventions, and Pydantic v2 API boundaries across Python 3.10–3.12+ services.