2.6 Creating and Configuring a Pipeline Run

With the pipeline imported, you are now ready to create a "run". A pipeline run is a single execution of the pipeline’s workflow. For this pipeline, you need to provide several parameters so it knows how to connect to the mock API and the Milvus database.

Procedure

  1. From the pipeline’s details page (where you landed after the import), click the Create run button in the top-right corner.

    The Create run button on the pipeline details page.

  2. This will open the "Create run" page. First, make sure the correct Pipeline and Pipeline version are selected. If you have completed the Event-Driven PDF Ingestion lab, you will see two pipeline definitions — select ServiceNow Ticket Ingestion.

  3. You will need give the run a name (Pipeline Run in the example below) and description (Simple API to RAG pipeline), but the most important part is the Pipeline parameters section at the bottom.

  4. You must fill in the following parameters to tell the pipeline how to connect to the other services running in your project. Copy the values below exactly.

    • api_endpoint: The internal URL of the mock API you deployed.

      http://mock-servicenow-api-svc.userX.svc.cluster.local:8080/api/v1/incidents?state=closed

    • collection_name: The name of the collection to be created in Milvus.

      servicenow_incidents

    • milvus_host: The internal service name for the Milvus database.

      vectordb-milvus.userX.svc.cluster.local

    • milvus_port: The gRPC port for the Milvus database.

      19530

  5. After filling in all the parameters, the form should look like this:

    The completed pipeline run configuration form with all parameters filled in.

  6. Click the Create Run button to launch the pipeline run.

You will be automatically redirected to a real-time graph view of your running pipeline. Let’s explore how to monitor it in the next section.

Summary

  • Configured runtime parameters (api_endpoint, milvus_host, milvus_port, collection_name) that wire the pipeline to the services in your namespace

  • Created a pipeline run that will fetch incidents, generate vector embeddings, and insert them into Milvus

  • All service endpoints use in-cluster DNS — the same pipeline definition works across namespaces simply by changing the parameters