capabilities.work

Create a Deployment and Run a Prediction Through It

A 4-step recipe with Replicate. We can help you get access.

Recipe · 4 steps

Create a Deployment and Run a Prediction Through It

with Replicate · Artificial Intelligence Management

Create a deployment for a model version and run a prediction through it.

  1. List the available hardware SKUs so the requested hardware can be validated before creating the deployment.
  2. Create the deployment, pinning the model version to the chosen hardware and instance bounds.
  3. Run a prediction through the newly created deployment using its owner and name.
  4. Retrieve the prediction state, repeating via the retry branch until the prediction reaches a terminal status.
What you'll need to have ready
  • apiToken: Replicate API token used as a Bearer credential.
  • name: The name of the deployment to create.
  • model: The full name of the model to deploy e.g. stability-ai/sdxl.
  • version: The 64-character model version ID to deploy.
  • hardware: The hardware SKU to run the model on (from hardware.list).
  • minInstances: The minimum number of instances for scaling.
  • maxInstances: The maximum number of instances for scaling.
  • input: The model's input as a JSON object for the prediction.
For your developer

Arazzo workflow replicate:deploy-and-predict, 4 operations: hardware.list → deployments.create → deployments.predictions.create → predictions.get. Specs on APIs.io.

More recipes with Replicate

Pick a Model from a Collection and PredictResolve Latest Version and PredictRun an Official Model and PollCreate Prediction and Poll Until CompleteRun a Prediction with Bounded Wait and CancelScale a Deployment and Run a PredictionSearch for a Model and Run a PredictionStart a Training and Poll Until CompleteFetch Webhook Secret and Run a Webhook Prediction