Skip to Content
TutorialsFirst RAG (prompts)

Tutorial: First RAG (prompts)

Add a knowledge collection, ingest a document, and use it from the Agentron assistant or from an agent.

Prerequisites

  • Agentron running
  • An embedding provider configured (e.g. Settings → Embedding: Ollama, OpenAI, or OpenRouter)
  • A small document (e.g. a text file or a URL the assistant can fetch)

Prompts to try

Open Chat and use the assistant where possible; some steps are easier in the Knowledge UI.

1. Create a collection (often via UI)

  • In the app: go to Knowledge, create a new collection (e.g. “Docs”). Choose encoding and embedding provider.
  • Or ask in chat: “How do I create a RAG collection?”: The assistant can explain and point you to Knowledge.

2. Add a document

  • In Knowledge: open your collection, add a document (upload a file or paste text / URL depending on your setup).
  • Or in chat: “Add this to my Docs collection: [paste a short paragraph].”: If the assistant has a tool or path to ingest text, it will use it.

3. Use knowledge in chat

  • “Using my Docs collection, what does it say about X?”: The assistant can use RAG retrieval (if enabled for chat) and answer from the ingested content.
  • “Search my knowledge for ‘Y’.”: Same idea; the assistant retrieves and summarizes.

4. Attach collection to an agent (via UI)

  • In Agents: edit an agent and set RAG collection to your collection. Runs of that agent can then use the collection for retrieval.

What’s going on

  • Collections hold documents that are chunked and embedded.
  • Embedding provider (Settings → Embedding) turns text into vectors.
  • Chat and agents can query the collection so answers are grounded in your data.

For full setup (encoding, vector store, document store), see the Knowledge section in the app and any Guides on RAG when added.

Next

Last updated on
First workflow (prompts)Tools