The GCP AutoML service allows us to train custom models and then run predictions against them. The Node-RED AutoML node allows us to execute a prediction request. It encapsulates the predict API documented here. On execution the msg.payload
is expected to have child nodes of:
- msg.payload.payload
- msg.payload.params (optional)
corresponding to the payload and params data requested by the underlying GCP API. The response is found in msg.payload
.
The configuration of the node requires:
Project
- The project containing the model.Location
- The region containing the model.Model ID
- The API key for the model (usually TCN…..).