bugfix when initializing with async aoss vector store

[Aydin-ab]

Description

@logan-markewich
Following up this issue [Bug]: Can't initialize OpensearchVectorClient through AWS endpoint in async mode #16746
16746

At the moment, we can't initialize a OpensSearchVectorClient using AOSS (amazon opensearch serverless) and/or an async connection. There are 2 problems here:

• Problem 1: If initializing with AOSS (async or not), the code will fail when fetching the version number because this requires a GET / call to the endpoint. This works on standard Opensearch but OpenSearch Serverless doesn't support that command.
• When using an async connection, the code will fail when getting or creating/refreshing the indices in the __init__(). This is because the code doesn't use the async client but only the "sync" one, this raises a TypeError due to a coroutine appearing somewhere.

Fixes # (issue)

• Problem 1: The version number is used only to check the eligibility of the efficient-kNN filtering. I couldn't find any other ways to check the version of the cluster so I decided to remove that efficient-kNN filtering feature completely when we're using AOSS (we check self.aoss). Ideally, we'd like to find a workaround to check the version so we can still use efficient-kNN filtering while using AOSS.
• Problem 2: I fix that with an extra except clause that catches a TypeError and retry with the async client this time.

New Package?

Had to import asyncio to run the async client but it's a python built-in anyway

Version Bump?

• Yes
• No

Type of Change

Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)

How Has This Been Tested?

Not sure how to test that, we'd need to simulate an endpoint ?

• I believe this change is already covered by existing unit tests

Suggested Checklist:

• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• I have added Google Colab support for the newly added notebooks.
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• I ran make format; make lint to appease the lint gods

[logan-markewich]

this is a little error prone -- a little safer to do

from llama_index.core.async_utils import asyncio_run

asyncio_run(...)

This will make sure any existing async loop is used

[logan-markewich]

just curious, there's no way to create an sync client from an async client, or vice-versa? This would solve a lot of headaches if possible

[Aydin-ab]

I guess you could if you dig into the kwargs and change it accordingly for each sync/async instantiation

from opensearchpy import AsyncHttpConnection, AWSV4SignerAsyncAuth, RequestsHttpConnection, AWSV4SignerAuth

# Sync parameters
kwargs['auth'] = AWSV4SignerAuth(credentials, region, 'aoss')
kwargs['connection_class'] = RequestsHttpConnection
self._os_client = os_client or self._get_opensearch_client(
    self._endpoint, **kwargs
)

# Async parameters
kwargs['auth'] = AWSV4SignerAsyncAuth(credentials, region, 'aoss')
kwargs['connection_class'] = AsyncHttpConnection
self._os_client = os_client or self._get_opensearch_client(
    self._endpoint, **kwargs
)

But you'd need the AWS credentials/region/service ('aoss') as additional inputs, seems too much

[logan-markewich]

ah yea, thats pretty messy. Too bad

[logan-markewich]

Looks like you'll need to run make lint

[Aydin-ab]

Looks like you'll need to run make lint

Sorry I don't know how that works :(