test: recategorize llm based astream_incremental & add versions with mocks

[planetf1]

Misc PR

Fixes #562

Type of PR

• Bug Fix
• New Feature
• Documentation
• Other

Description

• Link to Issue: test: test failed in CI, may be qualitative #562

The tests in test_astream_incremental.py were flaking in CI because they depended on live LLMs streaming responses. Due to the non-deterministic nature of token streaming (chunk arrival times and sizes), exact string matching assertions would fail intermittently.

To address this:

1. All tests in test/core/test_astream_incremental.py are now marked with @pytest.mark.qualitative. This follows the Mellea convention for tests dependent on varying live LLM interaction, ensuring they are skipped in CI but remain available for local execution during development.
2. A new test file, test/core/test_astream_mock.py, has been introduced. These tests use an asynchronous mock queue to deterministically test the incremental chunking logic of ModelOutputThunk without relying on a live LLM backend or network timing, guaranteeing they run reliably in all environments including CI.

Testing

• Tests added to the respective file if code was changed
• New code has 100% coverage if code as added
• Ensure existing tests and github automation passes (a maintainer will kick off the github automation when the rest of the PR is populated)

[github-actions]

The PR description has been updated. Please fill out the template for your PR to be reviewed.

[mergify]

Merge Protections

Your pull request matches the following merge protections and will not be merged until they are valid.

🟢 Enforce conventional commit

Wonderful, this rule succeeded.

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert|release)(?:\(.+\))?:

[planetf1]

My results (macOS):

> uv run pytest test/core/test_astream_mock.py -v
================== test session starts ===================
platform darwin -- Python 3.12.8, pytest-9.0.0, pluggy-1.6.0 -- /Users/jonesn/src/mellea-review/.venv/bin/python3
...
collected 6 items                                        
test/core/test_astream_mock.py::test_astream_returns_incremental_chunks PASSED [ 16%]
test/core/test_astream_mock.py::test_astream_multiple_calls_accumulate_correctly PASSED [ 33%]
test/core/test_astream_mock.py::test_astream_beginning_length_tracking PASSED [ 50%]
test/core/test_astream_mock.py::test_astream_empty_beginning PASSED [ 66%]
test/core/test_astream_mock.py::test_astream_computed_returns_full_value PASSED [ 83%]
test/core/test_astream_mock.py::test_astream_final_call_returns_full_value PASSED [100%]
====================== 6 passed in 3.81s =======================

[jakelorocco]

I like the addition of the mock tests. But can you please explain:

The tests in test_astream_incremental.py were flaking in CI because they depended on live LLMs streaming responses. Due to the non-deterministic nature of token streaming (chunk arrival times and sizes), exact string matching assertions would fail intermittently.

It looks like all the tests just match the values of chunks. I think these should be exact matches no matter the conditions since we aren't relying on the LLM to reproduce output; we are relying on us saving that non-deterministic output correctly?