Question 1

What is retrieval-augmented generation (RAG)?

Accepted Answer

RAG combines vector search with a language model. Your documents are split into chunks, embedded, and stored in a vector database. At query time the most relevant chunks are retrieved and passed to the model as context, so answers are grounded in your source material rather than the model's training data.

Question 2

What powers this demo?

Accepted Answer

FastAPI for the API layer, ChromaDB for vector storage, SentenceTransformers (all-MiniLM-L6-v2) for local embeddings, and the OpenAI Responses API for answer generation.

Question 3

Is my uploaded document shared with other users?

Accepted Answer

No. Each upload is tagged with your session id, and retrieval only returns chunks from your own session plus the shared public knowledge base.

AI Support Agent — RAG Demo

Session

Knowledge ingestion

Indexed sources

RAG chat