Loop
Loop is a model you define in the app, then update by sending outcomes to a train endpoint. It does not train on Lake gold.
Use it when each prediction can later be paired with what actually happened. A spreadsheet you already labeled is ML. A question about a piece of text, with answers you write, is Resolve.
Data Intelligence → Loop (/loop)
| Loop | ML | Forecast | |
|---|---|---|---|
| How it learns | Each successful /train/ call | One training job on gold | One training job on a time series |
| What you send later | Input fields, then the outcome | Feature columns | How many steps ahead |
- Define the fields
- Issue a key on the loop page
- Predict, then train
If you never call /train/, the model stays as it was when you created it.