“I need to choose between Chroma, pgvector, and Pinecone for a RAG system that will serve about 100 users and index around 10,000 document chunks. We’re already running PostgreSQL. Walk me through the trade-offs and give me a clear recommendation.”
“How does HNSW indexing work in a vector database, and why is it so much faster than brute-force similarity search? Explain it in plain English without maths notation.”
“I’m building a multi-tenant RAG application where each customer’s documents must be kept completely separate. Show me how to implement this using Chroma’s metadata filtering so that user A’s queries never retrieve user B’s documents.”
“Show me how to build a document ingestion pipeline in Python that detects when a source document has changed, removes its old chunks from a Chroma collection, re-chunks the new version, and re-indexes it.”
“What is hybrid search in the context of vector databases, and when would I use it over pure semantic search? Give me a concrete example where keyword matching matters alongside semantic similarity.”
