What problem does it solve? When a project brief or design document calls for a vector database for RAG-style retrieval, this Skill scaffolds the chosen store's client, ingestion, and query code consistently, so you never hand-write store-specific boilerplate or accidentally swap in a different database than the one specified. ## Core Features & Use Cases - Three supported stores: ChromaDB (local, on-disk persistence), FAISS (pure local, in-memory or saved index), and Qdrant Cloud (requires QDRANT_URL and QDRANT_API_KEY). - Consistent abstraction: Generates one small module (e.g. vector_store.py) exposing only upsert(chunks, embeddings, metadata) and query(embedding, top_k), so callers never touch the client SDK directly. - No mock mode: Qdrant Cloud connections are real network calls; missing credentials or failed connections are a hard stop, never a silent fallback to a local store. - Use Case: A teaching demo's design.md names Qdrant Cloud for retrieval. The Skill adds qdrant-client, documents the required env vars in .env.example, scaffolds the vector_store module, and documents the collection name and reset procedure in the data README. ## Quick Start Scaffold the vector store module for the database named in my design.md, including ingestion and query functions plus the required dependencies and environment variables.
