Skip to main content

Replicate Trigger

Replicate is a cloud platform for running machine learning models. This integration connects Replicate with Boltic so you can trigger workflows in real time when prediction lifecycle events occur — such as when a prediction starts, produces output, emits logs, or completes.

Replicate Account​

  • Select the Replicate account you want to connect.
  • If you do not have an existing account, add one using the Add New Account option before proceeding.

Available Resource​

Here is the available resource you can use with the Replicate trigger node:

Resource​

  • Webhook: A webhook is a real-time HTTP callback that sends prediction lifecycle event data to your workflow URL when a prediction changes state on Replicate.

Basic Operations​

Here are the basic operations you can perform with the Replicate webhook trigger:

Webhook Scope​

  • Prediction: Triggered when a prediction you create reaches the selected lifecycle event.

Prediction Events​

  • Prediction Started: Triggered when a prediction begins processing on the selected model version.
  • Prediction Completed: Triggered when a prediction finishes successfully, fails, or is canceled.
  • Prediction Output: Triggered each time a prediction emits an output chunk during streaming.
  • Prediction Logs: Triggered each time a prediction emits log lines.

Trigger Configuration​

  • Prediction Event: Select the lifecycle event that should trigger your workflow (Started, Completed, Output, or Logs).
  • Model Version: Enter the 64-character ID of the model version you want to run. Replicate will create a prediction using this version and route events to your workflow.
  • Input: Provide the model input as a JSON object. The required keys depend on the specific model version you selected.

Conclusion​

Replicate webhooks let you build real-time automations around AI model predictions. Once configured, your workflow will receive event payloads whenever the selected prediction lifecycle event occurs for the model version you specified.

If you need further assistance, please contact Boltic support at [email protected].