# Prediction

> Use task_type predict with current and previous scene context to predict likely future spatial states.

Source: https://fourechelon.com/docs/prediction · Machine-readable: https://fourechelon.com/openapi.json · Index: https://fourechelon.com/llms.txt

## task

predict

## question

Given where things were and are, where do they go next?

## required

| Field | Type | Description |
| --- | --- | --- |
| input.point_cloud_scene | object | Current scene state. |
| input.previous_point_cloud_scene | object | Prior state. Without it there is no motion history to extrapolate. |

## returns

| Field | Type | Description |
| --- | --- | --- |
| spatial_prediction.objects | Projected positions per tracked object. |
| spatial_prediction.confidence | Per-object confidence. |
| spatial_prediction.horizon | How far ahead the estimate reaches. |

## notes

Prediction needs two states. Sending only the current scene returns present-state reasoning, not a forecast.

## Example

```bash
curl https://fourechelon.com/api/v1/responses \
  -H "Authorization: Bearer $FOUR_ECHELON_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "rowe-1.0",
    "input": {
      "task_type": "predict",
      "previous_point_cloud_scene": { "point_clouds": [] },
      "point_cloud_scene": { "point_clouds": [] }
    }
  }'
```