RAG Agent
The RAG Agent is PrimePilot's general-purpose knowledge assistant. It answers user questions by retrieving relevant content from a knowledge base and generating a response grounded in that material.
It is included on Starter plans and is a common choice for chat widgets, full-page chat, and API deployments.
What it does
- Retrieves the most relevant passages from assigned knowledge sources
- Generates answers that reflect your documentation and data
- Can include source citations so users can verify answers
- Supports custom system prompts, models, and response style
When to use it
Use a RAG Agent when you need a flexible assistant that speaks from your own content: product docs, FAQs, internal policies, marketing copy, or any material you index in PrimePilot.
Until you attach knowledge sources, the agent behaves like a simple LLM proxy. Once sources are connected and processed, it becomes a full retrieval-augmented assistant.
Configuration
Configure each RAG Agent instance from the dashboard:
- Knowledge bases and sources: assign one or more knowledge bases or agent-specific sources. See Knowledge Sources.
- LLM settings: choose provider, model, temperature, and system prompt.
- Embedding and vector store: set how content is indexed and searched (Pro plans offer more options).
- Response style: for example professional or concise tone, and whether to show sources.
Typical usage
Add a RAG Agent to a squad and deploy via Full Page Chat, Floating Widget, Inline Widget, or API.