Python Development

Philosophy

Readable, concise code. Function names and type hints should make purpose clear without verbose docstrings.

Tooling

•Package manager: uv - new projects: uv init --python 3.14
•Linting/formatting: ruff
•Testing: pytest
•Logging: loguru

Preferred Libraries

Purpose	Library
CLI tools	`typer`
CLI output	`rich` (when needed)
HTTP clients	`httpx`
Web APIs	`fastapi` + `uvicorn`
DataFrames	`polars`
ORM	`sqlmodel` (Pydantic + SQLAlchemy)

Code Style

•Type hints on function signatures
•Small, focused functions
•pathlib for paths
•dataclasses or Pydantic for structured data
•SQLite for prototyping, Postgres for prod

Docstrings

Minimal but useful. Skip verbose Google/NumPy style. Answer: "what does this return in practice?"

python

def fetch_weather(city: str) -> dict[str, float]:
    """Get current weather from OpenWeatherMap API.

    Returns {"temp": 24.5, "humidity": 65.0, "wind_speed": 12.3}
    """

Include: what it does (if name isn't obvious), what's actually returned (not just the type), optional example for complex functions.

Google APIs

Use Google SDKs directly (gspread and similar wrappers are outdated/unmaintained).

bash

uv add google-api-python-client google-auth-httplib2 google-auth-oauthlib

Refs: Sheets quickstart, reading values

Local Testing

Expose local servers via Cloudflare tunnel:

bash

cloudflared tunnel --url http://localhost:8000

Deployment

Railway for FastAPI projects. Volumes required for persistent files (SQLite db, credentials.json).

Ref: Railway LLM docs