Upgrade redisvl and improve resource management

[tylerhutcherson]

• Updates RedisVL and other libraries to best ranges.
• Implements lifespan hooks for app startup and cleanup.
• Cleans up resources on shutdown.
• Fixes test fixtures for ideal loading and usage.

[tylerhutcherson]

Wierd -- when I implement the same updates to arxiv search I am seeing the warning/errors in the logs":

https://github.com/redis-developer/redis-arXiv-search/actions/runs/13512854236/job/37756259025?pr=32

I must be doing something different/wrong there. My eyes are tired! Will check back in the morning.

[abrookins]

Ah, yes. FastAPI's lifespan context manager doesn't run during tests if you use AsyncClient, only TestClient (and write your tests with sync Python). The logging I/O error is related to Pytest, which creates the ugly traceback, but the message means we have an async client instance that isn't getting cleaned up. That'll be true if the lifespan function never runs during tests (it doesn't, I checked).

There are some elaborate ways you can make the lifespan events fire during tests, but in our case, we can get away with an autouse Pytest fixture. Just pushed to your branch.

As for the Pytest logging issue, there isn't a great way to prevent that, but we can use another fixture to help. I pushed one up. But because we're logging within other people's projects as a library, it may be worth suppressing the logging message in our weakref finalizer.

[tylerhutcherson]

Ah, yes. FastAPI's lifespan context manager doesn't run during tests if you use AsyncClient, only TestClient (and write your tests with sync Python). The logging I/O error is related to Pytest, which creates the ugly traceback, but the message means we have an async client instance that isn't getting cleaned up. That'll be true if the lifespan function never runs during tests (it doesn't, I checked).

There are some elaborate ways you can make the lifespan events fire during tests, but in our case, we can get away with an autouse Pytest fixture. Just pushed to your branch.

As for the Pytest logging issue, there isn't a great way to prevent that, but we can use another fixture to help. I pushed one up. But because we're logging within other people's projects as a library, it may be worth suppressing the logging message in our weakref finalizer.

To get around the lifespan issue, I had used the asgi-lifespan pkg LifespanManager as a context manager around the httpx client and that manually would fire those lifespan events and cleanup on exit. It adds yet another dependency, but it is "clean".

Only thing tripping me up is that I see the logging error/warning for the arxiv demo but not this one.

[rbs333]

Looks good to me maybe we can rubber duck on the warning you're getting in arxiv later.

[tylerhutcherson]

Landed on using the LifespanManager class from asgi-lifespan because the hooks don't fire for pytest

[abrookins]

LGTM! 👍