Research RAG

Semantic search over embedded markdown in Neon harness_embeddings.

Prerequisites

• docs/rag/neon-doppler-setup.md completed
• Neon MCP enabled in Cursor
• Corpus ingested (embed-markdown.py or manual INSERT)

Process

1. Confirm config/rag.yaml or config/rag.template.yaml — corpus_id, top_k.
2. Embed the user question with Voyage (input_type="query") or use nearest MCP.
3. Run similarity SQL via Neon MCP: