Skip to main content

Datapoints

Datapoints are structured fields the agent can collect from users. They control validation, formatting, and deterministic data checks.

Datapoint Structure​

{
  "name": "Email Address",
  "slug": "email",
  "description": "Customer email",
  "type": "email",
  "validation": {
    "required": true
  },
  "ui": {
    "placeholder": "john@example.com",
    "examples": ["user@company.com"]
  }
}

Supported Types​

  • string
  • email
  • phone
  • number
  • boolean
  • date

Validation Examples​

{
  "validation": {
    "minLength": 5,
    "maxLength": 100,
    "pattern": "^[A-Z]{2}-\\d{4}$"
  }
}
{
  "validation": {
    "min": 0,
    "max": 1000,
    "integer": true
  }
}

Using Datapoints In Capabilities​

Collect datapoints in a collect step:

- collect: [email, phone, order-id]

Collected values are stored in variables and can be referenced later:

- done: "We will contact you at {{variables.email}}"

Tips​

  • Use clear descriptions to improve extraction accuracy
  • Keep slugs consistent across capabilities
  • Prefer optional fields for nice-to-have data