Module 11: Fleet Orchestration
Duration: 60 minutes Day: Day 3, Session 5
The hands-on lab for this module lives in the Hermes repository. The lab is designed as a team exercise for the live workshop. See the Solo Learner callout below if you are completing this course independently. Read the Concepts page first, then open the Hermes lab guide.
If you are completing this course independently (Udemy or self-paced), the team lab can be adapted: build all three track agents from Module 10 (A, B, C) sequentially, then wire them with a coordinator. You play the role of each domain specialist. The coordinator receives the cross-domain incident and delegates to whichever specialist agent covers that domain. This gives you the same architectural experience as the team lab — it just takes longer (allow 90 minutes total instead of 60).
What This Module Is About
One agent is useful. A team of agents is powerful.
Your Module 10 domain agent is excellent within its domain. But real incidents cross domain boundaries: a cost spike may trace to a Kubernetes deployment that caused EC2 scaling, which triggered RDS connection exhaustion. No single domain agent sees the full picture.
Fleet orchestration is the pattern for wiring specialized agents together: a coordinator agent receives a task, analyzes which domain it touches, delegates to the appropriate specialists, collects their diagnoses, and synthesizes a cross-domain response.
Learning Objectives
By the end of this module, you will be able to:
- Combine agents into a coordinated fleet with a manager (coordinator) agent and specialist (domain) agents
- Configure delegation between agents using the Hermes
delegate_taskmechanism - Explain three fleet patterns (round-robin, skill-based, hierarchical) and when each is appropriate
- Handle a cross-domain incident scenario where the coordinator synthesizes inputs from multiple specialists
Prerequisites
- Module 10 completed (you have at least one working domain agent)
- Hermes with
delegate_tasktool enabled - Team lab: Three team members, each with a different Module 10 track agent
Module Contents
| Section | Content | Time |
|---|---|---|
| Reading | Concepts: Fleet Patterns and Delegation | 15 min |
| Reading | Reference: Fleet Configuration and Coordinator Templates | 10 min |
| Lab | Fleet Orchestration — Team Exercise (Hermes repo) | 25 min |
| Quiz | Module 11 Assessment | 10 min |
| Exploratory | Stretch Projects | Optional |
Key Insight: Coordination Without Retraining
Fleet orchestration does not require a more capable AI model. It requires better architecture. A coordinator that delegates "Kubernetes pod analysis" to a specialist with a K8s skill and correct tool access will produce better results than a general-purpose agent that handles all domains at once — not because the coordinator is smarter, but because the specialist has the right context, tools, and procedural memory for its domain.
This is the microservices principle applied to agents: do one thing well, coordinate through well-defined interfaces.