RAG

RAG is a technique where an AI system first retrieves relevant documents or data, then uses that retrieved context to generate a more accurate and grounded response. It reduces hallucination by anchoring the model to real sources.

In operations, RAG is useful when an agent needs to answer questions based on project documents, past emails, or knowledge bases. Instead of relying on what the model was trained on, it looks up the specific information needed.

The OpsForEnergy agents use a lightweight form of RAG when composing the Ops Supervisor digest. It queries Supabase for the relevant week's activity and includes those records in the prompt, ensuring the summary is factually grounded.