Both RAG examples converted with embeddings regenerated

[borisno2]

Parent

#1129 — build spec 9 of the Prisma 8 architecture spec. Lands on the prisma-8 integration branch, not main.

What to build

rag-ollama-demo and rag-openai-chatbot run on the Dev database with pgvector installed by Prisma from the seeded space, embedding() on the native column with its index, embeddings regenerated on seed from source text, and search through nearest().

Acceptance criteria

• Each example's first pnpm dev installs pgvector without a hand step and seeds embeddings
• Semantic search returns ranked, scoped results with a non-sequential plan
• No JSON-to-vector conversion exists; the setup doc and install script are gone

Blocked by

• rag_search through nearest(); RAG docs to provisioning; install script deleted #1169
• Starter and starter-auth converted end to end with e2e green #1170

---

Generated by Claude Code