fix: MLPA disconnect causes Runtime Error

[FieldMarshallObvious]

What's Changed

• Removes manual await response.aclose() in cancel branch
• Suppresses runtime errors in httpx teardown
• Added tracking to capture dict to allow mocks to track how many times the stream was closed
• Added tests to verify that stream close call and validate that errors are not thrown

QA Path

Added two test cases, added a test case to ensure that the stream is called once per close and that Runtime error is not thrown.

Close Count

In this test we run the close count validation on the old logic versus the new logic

Old Logic:

New Logic:

Runtime error

In this test we validate that runtime errors are suppressed in the httpx teardown

Old Logic:

New Logic:

Related Tickets & Documents

AIPLAT-1124

[subpath]

I suggest to approach it with TDD : )

First make a test to reproduce the same error from the ticket, it should be failing - that would be RED
then you make a fix in completions.py and the test from step 1 should start passing
then refactor if needed

[FieldMarshallObvious]

I suggest to approach it with TDD : )

First make a test to reproduce the same error from the ticket, it should be failing - that would be RED then you make a fix in completions.py and the test from step 1 should start passing then refactor if needed

Added the tests you suggested and updated the PR message!

[subpath]

I suggest to approach it with TDD : )
First make a test to reproduce the same error from the ticket, it should be failing - that would be RED then you make a fix in completions.py and the test from step 1 should start passing then refactor if needed

Added the tests you suggested and updated the PR message!

Thank you ❤️

[subpath]

#198 (comment)

after than it's good to go, thank you ❤️