GenAI Playbook Repository¶
Read the playbook before you change or operate a live RAG or agent system. Each playbook tells you what to do, the steps to follow, how complex and risky it is, what goes wrong, what to alert on and what you must know first.
Status
Only playbooks marked published have been dry-run by a second engineer and signed off. Everything else is a draft outline: useful to read, not safe to follow on a live system.
- RAG playbooks: 47 playbooks in 12 packs
- Agentic AI playbooks: 53 playbooks in 12 packs
Playbook status¶
Status summary¶
| Status | Playbooks | Operational readiness |
|---|---|---|
| Published | 0 | Validated in sandbox and signed off. Safe for production use. |
| Validated | 0 | Validated in sandbox. Awaiting reviewer sign-off. |
| Draft | 100 | Draft outline. Review before use; not verified on live systems. |
| Stale | 0 | Validated >90 days ago without revalidation. Unsafe until re-tested. |
| Retired | 0 | Deprecated or superseded. Do not use. |
Pack status overview¶
| Pack | Topic | Owner | Total | Published | Validated | Draft | Stale |
|---|---|---|---|---|---|---|---|
| RAG-01 | Chunking: choose, benchmark and re-chunk safely | Yogeshwaran | 4 | 0 | 0 | 4 | 0 |
| RAG-02 | Retrieval evaluation and regression gates | Prasanna | 4 | 0 | 0 | 4 | 0 |
| RAG-03 | Index design, schema change and cut-over | Yogeshwaran | 2 | 0 | 0 | 2 | 0 |
| RAG-04 | Permission-aware retrieval and security trimming | Yogeshwaran | 4 | 0 | 0 | 4 | 0 |
| RAG-05 | Document ingestion: build, refresh and recover | Yogeshwaran | 4 | 0 | 0 | 4 | 0 |
| RAG-06 | Citations, grounding and no-answer handling | Deeparani | 5 | 0 | 0 | 5 | 0 |
| RAG-07 | RAG prompt templates and safe prompt changes | Deeparani | 4 | 0 | 0 | 4 | 0 |
| RAG-08 | Hybrid search and re-ranking | Deeparani | 4 | 0 | 0 | 4 | 0 |
| RAG-09 | Knowledge base setup and refresh | Yogeshwaran | 4 | 0 | 0 | 4 | 0 |
| RAG-10 | Multimodal, table and figure extraction | Yogeshwaran | 4 | 0 | 0 | 4 | 0 |
| RAG-11 | GraphRAG pattern kit | Deeparani | 4 | 0 | 0 | 4 | 0 |
| RAG-12 | RAG cost and latency tuning | Deeparani | 4 | 0 | 0 | 4 | 0 |
| AGT-01 | Agent templates: choose, build, run and upgrade | James | 4 | 0 | 0 | 4 | 0 |
| AGT-02 | Tool registry: define, secure, change and operate tools | James | 5 | 0 | 0 | 5 | 0 |
| AGT-03 | Planner patterns: choose, build, tune and set expectations | James | 4 | 0 | 0 | 4 | 0 |
| AGT-04 | Agent memory: design, set up, retain and erase | James | 5 | 0 | 0 | 5 | 0 |
| AGT-05 | Human-in-the-loop: approve, edit, reject, time out and audit | James | 5 | 0 | 0 | 5 | 0 |
| AGT-06 | Multi-agent collaboration: choose, build and operate topologies | James | 4 | 0 | 0 | 4 | 0 |
| AGT-07 | Agent evaluation & trajectory scoring | Prasanna | 4 | 0 | 0 | 4 | 0 |
| AGT-08 | Agent governance & permissioning | Prasanna | 5 | 0 | 0 | 5 | 0 |
| AGT-09 | Agent failure & recovery | Sivaraj | 4 | 0 | 0 | 4 | 0 |
| AGT-10 | MCP servers & tool interoperability | Sivaraj | 5 | 0 | 0 | 5 | 0 |
| AGT-11 | Agent cost & step-budget control | Sivaraj | 4 | 0 | 0 | 4 | 0 |
| AGT-12 | Agent simulation & scenario testing | Prasanna | 4 | 0 | 0 | 4 | 0 |
No playbooks are currently published for production use. Check the status summary above for the current validated, draft, stale and retired counts.
Stale playbooks (needs revalidation)¶
No playbooks are currently stale (re-validation required after 90 days).
Most urgent for live operations¶
- PB-AGT-02-02: Rotate credentials for MCP servers and tool connectors without downtime
- PB-AGT-02-05: Handle a tool dependency outage or throttling
- PB-AGT-05-03: Clear an approval queue backlog and handle timeouts
- PB-AGT-06-04: Investigate a multi-agent incident: ping-pong, conflicting writes or cost spike
- PB-AGT-07-02: Run the agent regression gate before a prompt, model or tool change
- PB-AGT-08-04: Respond to an agent acting outside its permission envelope
- PB-AGT-09-01: Stop and triage an agent stuck in a loop or burning tokens
- PB-AGT-09-02: Handle a tool or model outage during agent runs
- PB-AGT-10-03: Rotate MCP server credentials without breaking agents
- PB-AGT-11-03: Investigate an agent cost spike
- PB-RAG-01-03: Re-chunk and re-embed a live corpus without downtime
- PB-RAG-02-03: Triage 'the answers feel wrong': retrieval or generation?
- PB-RAG-04-03: Apply an urgent permission change or revocation
- PB-RAG-04-04: Permission leakage test and incident checklist
- PB-RAG-05-02: Recover from indexer and skillset failures and throttling
- PB-RAG-06-04: Investigate an ungrounded answer or wrong citation incident
- PB-RAG-07-01: Release a prompt change without regressions
- PB-RAG-07-02: Re-validate prompts on a model version upgrade or retirement
- PB-RAG-08-03: Diagnose and fix a relevance regression in production
- PB-RAG-09-01: Reindex an Azure AI Search index without downtime
- PB-RAG-09-02: Upgrade the embedding model and re-embed the corpus
- PB-RAG-12-01: Investigate a RAG cost spike
- PB-RAG-12-02: Diagnose and fix a latency regression end to end
Kinds of playbook¶
| Kind | Use it to |
|---|---|
| Runbook | Operate or change a live system: pre-checks, steps, verification, rollback, alerts |
| Decision guide | Choose between options: default, trade-offs, decision table, when to deviate |
| How-to | Build something with tested snippets and configs |
| Checklist | Check readiness before go-live or a change, and troubleshoot |