Playbooks

How to implement, paired with Blueprints for reference designs. One G.A.I.N subject maps to one blueprint and one playbook series.

Evaluation

Start with G.A.I.N Evaluation, then the Eval Blueprint and the Evaluation overview.

Group	Start here	Covers
Methods	Golden datasets	Datasets, synthetic, online, human review, LLM-as-judge
Planes	Input	Gates for planes ①-⑧

Agents

G.A.I.N Agents. Start at the Agents overview.

Group	Start here	Covers
Intent router	Overview	Route table, classifier, event bind, wire app, routing eval CI
Orchestration	Overview	Durable run pin, memory, autonomy shape, inference handoff
Manifests	Registry	Tool contract, pdp_action, version, pin

Also see Agents Blueprint. PEP: Runtime.

LLM

G.A.I.N LLM. Start at the LLM overview.

Playbook	Covers
Capability matrix	Approved endpoints per profile, task, data class, region
Gateway task routing	Plan vs synthesize vs classify; failover, cost, abstain
Canary promotion	Eval-gated model swap and rollback

Also see LLM Blueprint. Handoff from agents: Inference handoff.

Governance

G.A.I.N Governance. Start at the Governance overview.

Group	Start here	Covers
Operating	Overview	Inventory, obligation-to-control map, go-live evidence pack
Runtime	Overview	Foundation, assurance, five boundaries

Also see Governance blueprints (Operating · Runtime). Eval overlap: Action plane.

RAG

G.A.I.N RAG. Start at the RAG overview.

Playbook	Covers
RAG retrieval	Prefetch vs tool, scope, PEP pack, validation

Also see RAG Blueprint. Eval: Context plane.

MCP

G.A.I.N MCP. Start at the MCP overview. MCP is transport. Tool contracts live under Agents: Manifests.

Also see MCP Blueprint (transport) · Agents Blueprint (manifests). Eval: Tool plane.

Observability

G.A.I.N Observability. Start at Observability overview.

Playbook	Covers
Governance rules	Standards before scale
Business journeys	KPIs, events, outcome SLOs
Service signals	Golden signals, traces, SLIs
Infrastructure	Resource health linked to services
Correlation graph	KPI → trace → infra
Maturity	L0-L5 assessment
Operating model	RACI and cadence

Also see Observability Blueprint.