Using LLM Engine

Creating a Conversation

Once you have LLM Engine up and running, the first thing to do is create a Conversation. Since LLM Engine is a headless service, you can interact with the conversations API through any REST client such as Postman or directly through the OpenAPI swagger endpoint on your locally running server, at http://localhost:3000/v1/docs. If you wish to use a Front End to create and configure a Conversation, we recommend our open source Nextspace app.

The easiest way to create a conversation is to use the /conversations/from-type endpoint, which allows you to specify only the minimal properties needed to create a conversation that uses either our Event Assistant or Back Channel production agents.

Here is an example conversation that uses the Event Assistant:

{
    "name": "Where are all the aliens?",
    "topicId": "{{topicId}}",
    "type": "eventAssistant",
    "platforms": ["zoom"],
    "properties": {"zoomMeetingUrl": "{{ZOOM_MEETING_URL}}"}
}

Note: These agents require Zoom for transcription, so you must provide a valid Zoom Meeting URL. See our Zoom guide for more information.

Monitoring a Conversation

If you would like to monitor or run reports on agent responses and performance during a conversation, see our Monitoring and Reporting guide.

Experimenting with Agents

If you would like to experiment with an existing agent by modifying its prompts or changing its model, see our Experimentation guide.