manage-per-request-vs-application-wide-state

title: manage per-request vs. application-wide state

How to

• Application-wide: initialize in lifespan (runs once on startup) and store on the app; dispose on shutdown.
• Per-request: use async-generator dependencies; control reuse with NodeConfig(reuse=False) to ensure fresh instances.

Example

from ididi import NodeConfig
from sqlalchemy.ext.asyncio import AsyncEngine, AsyncConnection
from lihil import Route

async def get_conn(engine: AsyncEngine) -> AsyncConnection:
    async with engine.begin() as conn:
        yield conn

users = Route("/users", deps=[(get_conn, NodeConfig(reuse=False))])

References

• http/lifespan.md:1
• http/endpoint.md:1