Agent model settings

Model

The AI model that performs the text completion task. Here, reference a model from Models.

Response behavior

Response format

You can set up agents to either produce text or structured output. Select text when you want to create a chat client, or implement some other kind of text completion task.

Use json_schema when you want to produce structured output.

Tool calling is only available if the response format is set to text.

JSON Schema

Structured output configuration is only available if the response format is set to json_schema.

Open the editor by selecting the value help on the JSON schema field. The configuration must be a JSON object with two required top-level properties:

  • name — an identifier for the schema

  • schema — the JSON Schema definition describing the expected output structure

{
  "name": "support_ticket",
  "schema": {
    "type": "object",
    "properties": {
      "priority": {
        "type": "string",
        "enum": ["low", "medium", "high"],
        "description": "Urgency level of the support request"
      },
      "department": {
        "type": "string",
        "description": "Department responsible for handling the request"
      }
    },
    "required": ["priority", "department"],
    "additionalProperties": false
  }
}
Example structured output

You are building a customer service form, that has one main text area field. You do not want to burden the user with filling out structured data about their request, but you need to route the request to the right person.

To implement this, you decide to leverage an agent that shall output one of the categories of departmentA | departmentB | departmentC | unclear. To ensure that the output confirms to this schema, you select json_schema as the response format.

Temperature

The temperature controls the "creativity" of the output. The lower the temperature, the more "conventionally expected" the output is. Use a low temperature for tasks like classification. Using a low temperature may lead to the model repeating itself or the context.

The higher the temperature, the more "creative" and unexpected the results are. Use a higher temperature if your task is akin to creative writing.

Extended reasoning

Extended reasoning is only available when using a reasoning model.

Enabled

Toggle extended reasoning on or off.

Effort

Effort is only available for OpenAI reasoning models.

Control the reasoning effort the model applies before producing a response. Select one of the following levels:

  • low — fastest, fewer internal reasoning steps

  • medium — balanced

  • high — most thorough, more internal reasoning steps

Token budget

Token budget is only available for non-OpenAI reasoning models.

Set the maximum number of tokens the model may use for its internal reasoning steps. Enter an integer value.