[EPIC] Agent Orchestration - Persistent task management #95
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Overview
Infrastructure for persistent task management with Orchestrator service (`apps/orchestrator/`) as the execution backend.
Architectural Decision (2026-02-02 Update)
Architecture updated: Orchestrator is now a monorepo service at `apps/orchestrator/`, not external ClawdBot.
Delegation Model: Mosaic Stack integrates with Orchestrator service rather than building orchestration internally.
Scope
Mosaic Stack handles:
• Task definitions and CRUD API
• Database persistence (survives restarts)
• Progress UI and audit trail
• Dispatch to Orchestrator service backend
• Kill authority over all Orchestrator operations
• Failure handling (when Orchestrator reports failures)
Orchestrator service (`apps/orchestrator/`) handles:
• Agent spawning and lifecycle
• Health monitoring and heartbeats
• Automatic recovery
• Task coordination and scheduling
Phases
• Phase 1: Foundation (database schema, CRUD API)
• Phase 2: Orchestrator Integration (dispatcher, callbacks, kill endpoints)
• Phase 3: Failure Handling (retry logic, escalation)
• Phase 4: Observability (dashboard, kill controls)
Related
• #83 Federation EPIC (cross-instance coordination)
• #102 Gateway Integration (Orchestrator connection)
• ORCH-101 to ORCH-134 (Orchestrator implementation issues)