Journey Rag

It reads your plan from 02-plan.md, delegates the heavy lifting to rag-dev (which calls create_collection and ingest_document), then runs a smoke test with rag_query to make sure retrieval actually works. Gets skipped entirely if your plan doesn't include a knowledge base feature. The preflight check is smart, it won't proceed if your app_id is missing or docs are stale. One thing I appreciate is the anti-pattern callout about not ingesting huge document sets in one go, that's the kind of mistake you only make once.