AI Workflow Templates
RetrievalstarterHITL requiredupdated 2026-08-11

RAG Readiness Checker

Score whether a corpus is ready for RAG before you embed it: rights, freshness, structure, PII, and eval questions.

#rag#readiness#governance#checklist
View source →

Tools in the pipeline

  • api
    corpus-inventory
    List sources, owners, last-updated, access rights.
  • llm
    readiness-model
    Score each axis and name the blocking gaps.
  • human
    owner-review
    Confirm rights and PII before anything is embedded.
Stack: checklist · optional LLM

What it does

Answers 'should we even embed this?' before a RAG project starts. Scores rights, freshness, structure, PII, and whether you have gold questions to eval against. Listed as a free ARC Labs utility in the Master Blueprint.

Pipeline

  1. Inventory — sources, owners, dates, access.
  2. Score — rights, freshness, structure, PII, eval coverage (0-100).
  3. Blockers — name what would make retrieval unsafe or useless.
  4. Next — the smallest set of fixes before ingest.
  5. Gate — do not embed until blockers are cleared or explicitly accepted.

Prompt skeleton

Score this corpus for RAG readiness on five axes (0-20 each):
rights, freshness, structure, PII handling, eval coverage.
Return JSON: { score, grade, axes, blockers[], next[] }.
Do not recommend embedding if rights or PII are unknown.

Guardrails

Unknown PII or unknown rights is a blocker, not a warning. Do not treat a shared-drive dump as a knowledge base. Human accepts residual risk in writing.

Structural dry-run

No model call. Validates required fields only.

Prefills the example payload for Shared drive dump. Edit it, then run.

Related