Definition
Full definition of rag
Standard LLM call: user asks a question, model answers from training data. RAG: user asks a question, you first retrieve relevant docs (from a vector store), pass those docs as context, model answers from them. RAG dramatically reduces hallucination on domain-specific questions and lets you keep proprietary knowledge out of the model. Tiny Command makes RAG flows two steps: query vector store, pass results into the AI step.
In practice
RAG examples
RAG flow
User question → embed → search Pinecone → top-3 docs → prompt Claude with docs as context → grounded answer
Used by
Apps that exemplify rag
See rag in action across real integrations.
FAQ
Common questions about rag
Do I need a vector database for RAG?
For >1000 docs, yes. For small corpora, a simple keyword search or even loading all docs into the prompt works.
Can RAG fully prevent hallucination?
No — even with retrieved context, models sometimes invent details. Instruct the model: 'answer only from the sources; say I don't know if not found'.
