Your First Prediction

This guide walks you through making your first prediction with the Kanva API.

Understanding the Prediction Endpoint

The prediction endpoint accepts input features and returns model predictions:

POST /api/v1/projects/{id}/predict

Input data is sent as a JSON object where each feature is an array of values:

{
  "input": {
    "date": ["2024-01-15"],
    "category": ["A"],
    "value": [123.45]
  }
}

tip

Each feature value must be wrapped in an array, even for single predictions. This format supports batch predictions.

The API supports two modes:

Mode	Query Parameter	Response
Sync	`?sync=true`	Waits for result
Async	(default)	Returns job ID

For quick predictions (under 60 seconds), use synchronous mode. For longer-running predictions, use async mode with polling or webhooks.

{
  "success": true,
  "data": {
    "predictions": ["predicted_value"],
    "probabilities": [0.85, 0.15]  // classification only
  }
}

If something goes wrong, you'll receive an error response:

{
  "success": false,
  "error": "Invalid input: missing required feature 'date'"
}

See Error Handling for a complete list of error codes.