loading
loading
Retrieval-Augmented Generation — grounding model answers in fetched documents.
Instead of relying on what was baked into the model during training, RAG retrieves relevant chunks from a vector database at inference time and injects them into the prompt. This allows the model to answer questions about recent events, private data, or domain-specific knowledge it was never trained on. The quality of retrieval (chunk size, embedding model, similarity threshold) heavily influences answer quality.
Plainly
Think of RAG as a labeled box where an app keeps important things. Retrieval-Augmented Generation — grounding model answers in fetched documents.
In practice
Use it when a feature reads, writes, migrates, validates, or audits stored information. In practice, define the owner, input, output, and failure mode before you rely on it.