connect-raj/rag ? reverse-engineered prompt

Build me a simple JavaScript app that can take a set of documents, break them into chunks, turn those chunks into embeddings, and use a language model to answer questions from that content.

I want it to feel like a basic retrieval augmented generation demo, where I can drop files into a data folder, run one command to ingest them, and then ask questions in a small command line flow or through a single entry script. Make sure the pieces are separated in a sensible way, with clear modules for chunking, data ingestion, embeddings, and the model call.

Keep it easy to run locally, use environment variables for any API keys, and include a straightforward example data set so I can test it right away. If anything needs current docs or best practices, look them up online while you build it.