Ideas: Phase 4.3 — Production Deployment as the right next step after AgentMesh

[web3guru888]

The Question

Phase 4.2 (AgentMesh) is now four components strong:

• AgentMesh — central coordinator (Add AgentMesh coordinator: multi-agent task distribution and result aggregation #147)
• AgentDiscovery — service registry + heartbeat state machine (Add AgentDiscovery: service registry and health tracking for AgentMesh #150)
• MeshTaskQueue — priority queue with dead-letter handling (Add MeshTaskQueue: persistent task queue with priority, retry, and dead-letter handling for AgentMesh #154)
• MeshResultAggregator — four aggregation strategies + reliability feedback (Add MeshResultAggregator: collect, merge, and score results from distributed AgentMesh tasks #159)

So: what is Phase 4.3?

Why Production Deployment First

The architecture is complete on paper but not runnable end-to-end by a new contributor. git clone && python -m asi_build does not work yet. Before adding more research features, we should answer:

"Can someone actually run this?"

A deployable system:

1. Forces integration — wiring 29 modules together surfaces bugs that unit tests miss
2. Gives real performance numbers — CognitiveCycle tick latency, queue depth, memory footprint under actual load
3. Unlocks contributors — people who want to experiment but can't navigate 556 files can docker compose up
4. Enables the safety story — a policy engine is meaningless without a running system to govern

Proposed Phase 4.3 Scope

Sub-milestone	Deliverable	Issue
4.3.1	Docker Compose stack (FastAPI + Memgraph + Redis)	#164
4.3.2	/health, /ready, /metrics endpoints (backed by CycleFaultSummary)	#164
4.3.3	Prometheus metrics for CognitiveCycle, AgentMesh, Blackboard	#164
4.3.4	Kubernetes Helm chart	#164
4.3.5	GitHub Actions CI smoke test (docker compose + curl /health)	#164

What Phase 4.4 Would Look Like

Once the system is runnable in production:

• Agent Policy Engine (Option A from Ideas: Phase 4.3 — What should come after multi-agent orchestration? Policy engine, emergent coordination, persistent memory, or production deployment? #163) — now there are real dispatch logs to govern
• Persistent Agent Memory (Option C from Ideas: Phase 4.3 — What should come after multi-agent orchestration? Policy engine, emergent coordination, persistent memory, or production deployment? #163) — now there is actual per-agent performance data to cache
• Emergent Coordination (Option B from Ideas: Phase 4.3 — What should come after multi-agent orchestration? Policy engine, emergent coordination, persistent memory, or production deployment? #163) — now we can measure whether it outperforms central dispatch

Concrete Design Questions

1. Should docker compose include Memgraph, or should graph features be optional (env flag MEMGRAPH_ENABLED=false)?
2. Should Prometheus metrics be exposed as a sidecar (prometheus-client in a separate process) or embedded in the existing FastAPI server?
3. Is a Helm chart needed for Phase 4.3, or is Docker Compose sufficient to prove deployability?
4. Should the /ready probe require all 29 modules or a configurable quorum (default: 20/29)?

Full scope: #164