feat(test/perf): make performance_benchmarks.performance_test.py runnable — 3 pre-existing breakage sites (#7041 follow-up) #7132

Discovery during PR #7131 (#7041) — pre-existing benchmark file breakage

PR #7131 migrated performance_benchmarks.performance_test.py off the deleted LLMInterface and onto LLMService. While doing so, three pre-existing breakage sites surfaced that predate #3185 — the file has not been runnable in CI for a long time. Filing for tracked cleanup.

Specific gaps

1. enhanced_memory_manager module missing (line 18)

from enhanced_memory_manager import EnhancedMemoryManager

The repo only has enhanced_memory_manager_async.py — the synchronous module was removed at some point (likely the async migration). The import fails at module load with ModuleNotFoundError.

2. knowledge_base import chain broken (line 20)

from knowledge_base import KnowledgeBase

knowledge_base.py exists, but importing it raises ModuleNotFoundError: No module named 'models.task_context' — a transitive dep that's missing.

3. Non-standard filename → not pytest-collected

performance_benchmarks.performance_test.py doesn't match pytest's default test_*.py / *_test.py glob. Even if (1) and (2) were fixed, the file wouldn't run on CI without explicit --collect-only or a config addition.

Acceptance criteria

Decide between:

A. Resurrect — fix the 3 gaps:

• Migrate EnhancedMemoryManager import to enhanced_memory_manager_async (or its successor); update the test class accordingly
• Fix or stub models.task_context so knowledge_base.py imports cleanly in test contexts
• Either rename the file to performance_benchmarks_test.py (pytest-collected by default) OR add an explicit pytest.ini rule for the .performance_test.py extension AND mark with @pytest.mark.benchmark for opt-in
• Verify benchmarks actually run with a real Ollama instance and produce meaningful numbers

B. Retire — delete the file after confirming with the project maintainer that benchmark coverage is reproducible elsewhere (or filed as a separate "build new benchmark suite" issue). Per project policy [Never delete code — wire it in (ABSOLUTE)] this is the less preferred path.

The runtime-correctness layer of (A) is significant — at minimum it requires:

• Ollama running with llama3.2:1b and nomic-embed-text available
• Knowledge base initialized
• Memory manager initialized

That's effectively a "benchmark e2e suite" effort, separate from PR #7131's import cleanup.

Found via

• PR fix(test/perf): migrate performance_benchmarks.performance_test.py off LLMInterface (#7041) #7131 (discovery(test): performance_benchmarks.performance_test.py has stale LLMInterface — migrate or retire (#6983 / #3185 follow-up) #7041) sweep — py_compile passed but import of the patched modules failed during runtime mock-validation.
• Related: bug(tests): 48 mock paths target nonexistent 'src.*' package — patches silently no-op #6987 (the src.* mock paths bug) — already filed; its scope shrinks by 16 sites after fix(test/perf): migrate performance_benchmarks.performance_test.py off LLMInterface (#7041) #7131 lands.

Severity

P3 — file isn't pytest-collected and isn't blocking any CI run. Cleanup tracker only; no user-visible impact.