Rag Dev

This is a clean two-tool wrapper around Butterbase's pgvector RAG pipeline. You create collections with fixed chunking params, ingest text or files (PDF, DOCX, etc.) asynchronously, poll until embeddings are ready, then query with optional LLM synthesis. The mental model is simple: collections hold documents, documents become chunks, queries do cosine similarity search. Access modes let you go from private (service-key only) to shared (any authenticated user) to custom RLS. Metadata filtering is exact-match key-value, so design your tags upfront. The chunk_size and chunk_overlap are immutable once set, which feels restrictive but forces you to think through your chunking strategy before you load 500 PDFs. Good for support bots, versioned docs, or multi-tenant knowledge bases where you need semantic search without reinventing embeddings infrastructure.

Install to Claude Code

npx -y skills add butterbase-ai/butterbase-skills --skill rag-dev --agent claude-code