Router Playbooks
Playbooks · Router overview · Intent router (Plane ①) →
Implementation guides for the Router Blueprint. The insight and design guides explain why routing belongs at ingress and how to design it. These playbooks are the how for each plane.
THE CLAIM
Route before the loop. Plane ① picks the governed path; Plane ② runs the agent loop; Plane ③ selects the model at the gateway. Each plane has its own contracts, eval gates, and ownership.
Three planes
| Plane | Overview | What you build |
|---|---|---|
| Intent router (Plane ①) | Open → | Route table, layered classifier, agentic app wiring, routing eval CI |
| Orchestration (Plane ②) | PGAR Runtime | Agent loop, PEP/PDP, manifests — see PGAR playbooks |
| Model routing (Plane ③) | G.A.I.N LLM | Gateway profiles, failover, cost caps (dedicated playbooks planned) |
Recommended path
- Intent router (Plane ①): route table → eval CI in parallel → classifier → wire app
- PGAR Runtime (Plane ②): when manifests differ per turn, wire the agentic app into the governed loop
- Model routing (Plane ③): apply
model_profilefrom the route row at the LLM gateway
Bridge reading: What Is an Intent Router · How to Design an Intent Router · PGAR Agentic app · Eval Plane ①: Input.
Intent router playbooks at a glance
| Playbook | One-line purpose |
|---|---|
| Route table lifecycle | Versioned route contracts, storage patterns, promote and rollback |
| Layered classifier | Eligible routes, rules, classifier, LLM fallback, safety |
| Wire agentic app | Manifest load, outcomes UX, PGAR handoff |
| Routing eval CI | Golden set, release gates, incident replay |
Who should read what
| Role | Start with | Then |
|---|---|---|
| AI platform | Intent router overview, Route table lifecycle | Wire agentic app, PGAR Agentic app |
| Domain squad | Route table lifecycle | Routing eval CI |
| Governance | Routing eval CI | Eval Input plane |