Greptile Summary

This PR stabilizes the previously-failing test_goal_extraction_batch.py by updating its patching strategy from the removed llm_mini attribute to the current get_llm('goals') call path, and by adding a snapshot-and-restore mechanism for sys.modules so that temporary stubs don't leak into neighboring tests.

• Test patching strategy updated: all 18 test cases now use patch.object(goals_module, \"get_llm\", ...) instead of patch(\"utils.llm.goals.llm_mini\", ...), matching the current production code in utils/llm/goals.py.
• Module isolation hardened: a _restore_stub_modules() function captures the pre-stub state of sys.modules and parent-package attributes at load time, then restores them after the one-time isolated import of utils.llm.goals; the live module reference is preserved in _goals_module for use throughout the test session.
• test.sh updated: the file is added to the CI test script in the correct position between the KG-sanitization and listen-pipeline tests.

Confidence Score: 4/5

Safe to merge. The changes are confined to a test file and the CI runner script; no production code is touched.

The module-isolation machinery is well-reasoned and the 18 tests correctly verify the batched get_llm call path. The restore logic handles parent-attribute cleanup carefully. The three observations are quality improvements — none cause test failures today, but the incomplete _RESTORED_MODULES coverage and partial mock-reset scope could create subtle ordering sensitivity as the test suite grows.

The restore coverage in test_goal_extraction_batch.py (lines 55–65) is worth a second look to decide whether utils.llm and utils should be included in _RESTORED_MODULES.

Important Files Changed

Sequence Diagram

sequenceDiagram
    participant TF as test_goal_extraction_batch.py
    participant SM as sys.modules
    participant GM as utils.llm.goals

    TF->>SM: snapshot saved_modules and saved_parent_attrs
    TF->>SM: "install stubs for database.* utils.llm.clients utils.llm.usage_tracker"
    TF->>GM: importlib.import_module
    Note over GM: binds get_llm = stub, track_usage = stub, goals_db = stub
    GM-->>TF: _goals_module reference saved
    TF->>SM: _restore_stub_modules removes stubs and restores originals

    loop each test
        TF->>TF: reset_mocks resets get_user_goals and update_goal_progress
        TF->>GM: patch.object goals_module.get_llm with MagicMock returning mock_llm
        GM->>GM: extract_and_update_goal_progress calls goals_db.get_user_goals
        GM->>GM: calls get_llm via patched mock_llm.invoke
        GM-->>TF: result dict
    end

_{Reviews (1): Last reviewed commit: "Stabilize goal extraction batch tests" | Re-trigger Greptile}