RAG Playbooks
Playbooks · RAG overview · RAG retrieval →
Implementation guides for G.A.I.N RAG. The RAG Blueprint is the reference design. These playbooks are the how: who starts retrieve, which corpus is in scope, and how the pack is audited.
PEP still gates every retrieve. Enforcement lives in Runtime. This series owns the retrieval store and the pack, not the verdict engine.
THE CLAIM
Retrieval is a governed action, not a database query. The route declares retrieval. The app fetches. The LLM does not pick the corpus.
Playbooks
| Playbook | One-line purpose |
|---|---|
| RAG retrieval | Prefetch vs tool mode, scope, PEP pack, validation, worked example |
Recommended path
- G.A.I.N RAG: principles
- RAG Blueprint: reference design
- RAG retrieval: implement mode, scope, and the pack
- Eval Plane ③ Context: recall, scope leaks, abstention
- Runtime: PEP/PDP on the retrieve action
Who should read what
| Role | Start with | Then |
|---|---|---|
| RAG / knowledge teams | RAG retrieval | G.A.I.N RAG, Eval Context |
| AI platform | RAG retrieval | Route contract, PEP enforcement |
| Governance | Runtime | Retrieval scope scenarios in policy tests |