# Phase 5: Test Plan
This plan summarizes test coverage for the Memori-inspired features and the suggested verifications.
## Unit Tests
- Schema & Migration
- `tests/unit/test_schema_v2.py` – creates v2 tables
- `tests/unit/test_migration.py` – dry run and migration, version checks
- Extraction Cascade & Persistence
- `tests/unit/test_entity_extraction.py` – cascade success ordering
- `tests/unit/test_extraction_retry.py` – timeouts/retries per provider
- `tests/unit/test_persistence.py` – v2 inserts for conversations/entities/relationships
- Conscious Agent
- `tests/unit/test_conscious_agent.py` – promotions and demotions
- Filesystem Integration
- `tests/unit/test_fs_dedupe.py` – persistent dedupe behavior
## Integration Tests
- Filesystem-triggered extraction path (tool-level):
- `tests/integration/test_filesystem_extraction.py` – tool persists v2 row with activity-weighted importance
## Manual & Smoke Tests
- Run the server with feature flags defaulted to ON; confirm extraction, access logging, agent tools, and v2 schema behave as expected. Then toggle flags OFF via env to verify graceful degradation with pattern fallback and no side-effects.
- Enable v2 and run `migration_status`, `trigger_migration(dry_run=true)`, then backup+migrate
- Verify access logging by running searches and checking `access_log_stats`
- Start the Conscious Agent and verify a promotion/demotion run using `force_conscious_analysis`
- Enable filesystem extraction and edit a file in a monitored project; verify a new v2 entry
## Coverage & Performance
- Target 85% coverage overall (progressive)
- Latency/Performance: collect extraction `extraction_time_ms` values at call sites; aggregate externally
