Integration Service activities

Description​

• The auto-generated schema approach is unavailable for your model type.
• You need full control over the JSON payload sent to the endpoint.
• You are integrating with a custom or non-standard Databricks serving endpoint.

How to use the activity​

1. Add a service task element to the canvas and open the task's Properties panel.

2. In the Implementation section, from the Action dropdown list, select Start and wait for external agent.

3. Select the Databricks Agent connector.

4. Select an existing connection or create a new one. For more information, see Databricks Agent authentication.

5. From Activity, select Query a Serving Endpoint (Manual).

6. From Serving Endpoint, select the serving endpoint to invoke.

7. In the Input field, enter the full JSON payload to send to the endpoint. The expected payload structure depends on your model — refer to its documentation.

   For example, for a chat-based model, the payload might look like:

   assignment

   {
     "messages": [
       {
         "role": "user",
         "content": "What is the capital of France?"
       }
     ]
   }
   {
     "messages": [
       {
         "role": "user",
         "content": "What is the capital of France?"
       }
     ]
   }
   

8. Connect the start event to the service task, and the service task to an end event node on the canvas.

9. Select Debug to run the process. After a successful run, review the global variables and look for the response from the service task.

Troubleshooting and Tuning​

• The required JSON fields and their types.
• The structure of the response object returned by the endpoint.