Fix catchup prompt loading all entries into memory

[cmeans]

Summary

• The catchup prompt was calling get_knowledge() and get_active_alerts() without a since parameter, loading all entries into Python and filtering in-memory
• Both methods already accept a since parameter that pushes the filter to SQL — this PR passes it through
• Prevents linear performance degradation as the knowledge store grows

QA

Prerequisites

• pip install -e ".[dev]"
• Deploy to test instance on alternate port (AWARENESS_PORT=8421)

Manual tests (via MCP tools)

1. • Seed entries at different ages

   remember(source="qa-test", description="Old entry created well before window", tags=["qa-catchup"])
   

   Wait a moment, then:

   remember(source="qa-test", description="Recent entry created within window", tags=["qa-catchup"])
   

   Expected: Two entries exist in the store with different updated timestamps

2. • Verify catchup filters correctly
     Invoke the catchup prompt with a small time window (e.g., hours=1).
     Expected: Only entries updated within the last hour appear. The old entry (if created more than 1 hour ago) should not appear. If both are recent, both appear — the key is that the filtering happens server-side (verify via query logs or by checking that entries outside the window are excluded).
3. • Verify catchup with default window
     Invoke the catchup prompt with default parameters (no hours argument).
     Expected: Returns entries updated in the last 24 hours, grouped by source, with [new] or [updated] markers.
4. • Verify empty catchup
     Invoke the catchup prompt with hours=0.
     Expected: Returns "Nothing changed. You're up to date." since no entries were updated in the last 0 hours.
5. • Cleanup

   delete_entry(source="qa-test", tags=["qa-catchup"], confirm=true)
   

   Expected: Test entries removed

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[cmeans]

QA Review — PR #58: Fix catchup prompt loading all entries into memory

Automated Tests

• pytest: 333/333 passed
• ruff: 0 violations
• mypy: 0 errors

Code Review

Clean, minimal fix. since parameter correctly pushed to SQL for both get_knowledge() and get_active_alerts(). Eliminates O(n) in-memory filtering. No issues.

Manual QA (5/5 steps passed)

1. ✅ Seed entries at different ages

   • Created two entries with 2s gap — both confirmed in store with different updated timestamps

2. ✅ Catchup filters correctly (hours=1)

   • Returned 18 entries all within 1-hour window, grouped by source with [new]/[updated] markers

3. ✅ Catchup with default window (24 hours)

   • Returned 18 entries, all marked [new], grouped by source. Correct format.

4. ✅ Empty catchup (hours=0)

   • Output: "Nothing changed. You're up to date." ✓

5. ✅ Cleanup

   • delete_entry(source="qa-test", tags=["qa-catchup"], confirm=true) → trashed 2 entries ✓

Findings

None. Clean fix, all tests pass.