FastAPI: settings with pydantic-settings

Use pydantic-settings (BaseSettings) for deployment configuration (data directories, database URLs, executor sizing) and wire values into FastAPI’s lifespan alongside pydantable.fastapi.executor_lifespan.

Install

pip install "pydantable[fastapi]" pydantic-settings

pydantic-settings is not a runtime dependency of pydantable; add it in your service.

Recipe

from contextlib import asynccontextmanager

from fastapi import FastAPI
from pydantic_settings import BaseSettings, SettingsConfigDict

from pydantable.fastapi import executor_lifespan


class Settings(BaseSettings):
    model_config = SettingsConfigDict(env_prefix="MYAPP_")

    data_dir: str = "/var/data"
    sql_url: str = "sqlite:///./app.db"
    executor_max_workers: int | None = None


settings = Settings()


@asynccontextmanager
async def lifespan(app: FastAPI):
    async with executor_lifespan(
        app,
        max_workers=settings.executor_max_workers,
        thread_name_prefix="pydantable",
    ):
        yield


app = FastAPI(lifespan=lifespan)

Read paths and URLs from Settings in route dependencies or services; pass executor_max_workers (or None for the default pool size) to match worker capacity.

Example environment (with MYAPP_ prefix from the recipe above):

export MYAPP_DATA_DIR=/data/incoming
export MYAPP_SQL_URL=postgresql+psycopg://user:pass@db:5432/app
export MYAPP_EXECUTOR_MAX_WORKERS=8

Tune executor_max_workers from CPU count, expected concurrent long-running acollect / eager pydantable I/O calls, and process memory — not “as high as possible.”

Pitfalls

• Secrets: use env vars or a secret manager; do not commit credentials.
• Executor size: too many threads vs. CPU cores can hurt latency; tune per deployment.

See also

• GOLDEN_PATH_FASTAPI
• FASTAPI