profit
|
c96919fe35
|
Implement real auto-recovery with handoff chain
Orchestrator changes:
- Add dumpAgentHandoff() to dump proposals/analysis before abort
- Add loadRecoveryContext() to load inherited context on recovery runs
- Add preseedBlackboard() to pre-seed inherited proposals
- Force-spawn GAMMA immediately on recovery runs
- Track isRecoveryRun, recoveryAttempt, inheritedContext, forceGamma
Server changes:
- Update recordConsensusFailure() to read orchestrator handoff JSON
- Add collectFromBlackboard() helper as fallback
- Update triggerAutoRecovery() with comprehensive context passing
- Store inherited_handoff reference for recovery pipelines
- Track retry_count, abort_reason, handoff_ref in recovery:* keys
- Add recovery badge and prior pipeline link in UI
Test coverage:
- test_auto_recovery.py: 6 unit tests
- test_e2e_auto_recovery.py: 5 E2E tests (handoff dump, recovery
pipeline creation, inherited context, retry tracking, status update)
Redis tracking keys:
- handoff:{pipeline_id}:agents - orchestrator dumps proposals here
- handoff:{recovery_id}:inherited - recovery pipeline inherits from
- recovery:{pipeline_id} - retry_count, abort_reason, handoff_ref
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
|
2026-01-24 19:39:52 -05:00 |
|