Skip to main content

RAG Agent Setup

RAG (Retrieval-Augmented Generation) agents use your knowledge base to provide accurate, grounded responses. They do not require fine-tuning but can be used alongside it in other modes.

🧠 Overview​

RAG agents retrieve relevant Q&A content at runtime, grounding responses in your indexed knowledge base.

RAG Form

🧾 Form Fields​

Base Model​

Select the foundational model used to run retrieval-augmented responses (e.g., GPT-4.1 mini).

Enable Indexing for RAG Grounded Responses​

This toggle must be enabled to create a RAG agent.

Once enabled:

  • Your Q&A pairs are semantically indexed.
  • The agent fetches relevant answers during inference.
  • Responses are grounded and source-traceable.

⚠️ Important: Without enabling RAG indexing, the Create Agent button will remain disabled.

💡 Notes​

  • RAG can be used independently or later combined with fine-tuned models.
  • RAG-enabled agents automatically stay up-to-date as knowledge changes.
  • Runtime usage incurs a small retrieval overhead, typically ~2× standard model calls.