test: test failed in CI, may be qualitative

[ajbozarth]

A recent CI failed on an unrelated test which looks to be an unmarked qualitative test. We should mark it qualitative and check if other tests added with it should also be marked

https://github.com/generative-computing/mellea/actions/runs/22458644040/job/65046606449#step:13:824

[ajbozarth]

opening this since I'm in the middle of other work and don't want to pollute my work with separate bug fixes. I'll come back and fix this later if no one else does

[psschwei]

It looks like the test passed on 3.11 and 3.12, and it also passed locally when I tested with 3.10... could this possibly just be a flake?

[planetf1]

Looking at the 3.12 The test test_astream_final_call_returns_full_value in test/core/test_astream_incremental.py failed. The other tests in this file as similar

Marking as qualitative would be an appropriate first step - but in fact the test is really checking parsing/chunks are generated, not llm output, so the 'proper' fix is to rewrite it using a mock (I wonder if there's other tests a similar approach should be taken with...)

[planetf1]

I updated the tags which should be a sensible, safe fix.

However I do think we can improve the tests to more accurately test the mechanism rather than the llm. Rather than replace, they've been added as new tests that do not have the qualitative marker (so will run on a normal local test & in CI)