Creating Your First Pipeline

Build a deployment automation workflow step-by-step

This tutorial guides you through creating a production-ready deployment automation pipeline using the visual editor.

What You’ll Build

A deployment workflow that:

1. Validates cluster health
2. Creates or updates a deployment
3. Configures autoscaling
4. Sets deployment status as output

Prerequisites

• Access to Bud AI Foundry
• At least one registered cluster
• A model ready for deployment

Step 1: Create the Pipeline

1. Navigate to Pipelines in the sidebar
2. Click + Pipeline
3. Configure:
   • Name: “Deploy Model”
   • Description: “Automated model deployment with health check”
4. Click Create

The visual editor opens with an empty canvas.

Step 2: Add Cluster Health Check

Let’s start by verifying the cluster is ready:

1. From Actions panel → Cluster Operations
2. Drag Cluster Health Check onto canvas
3. This becomes your trigger node
4. Configure:
   • Cluster ID: params.cluster_id
   • Timeout: 30 seconds
5. Click Save

Using params.cluster_id makes the pipeline reusable across clusters.

Step 3: Add Deployment Creation

Now add the deployment step:

1. Drag a Transform action onto the canvas
2. Connect Cluster Health Check → Transform
3. Rename to “Create Deployment”
4. Configure:
   • Operation: deployment_create
   • Model ID: params.model_id
   • Deployment Name: params.deployment_name
   • Cluster: params.cluster_id
5. Click Save

Step 4: Configure Autoscaling

Add autoscaling configuration:

1. Drag another Transform action
2. Connect Create Deployment → this new Transform
3. Rename to “Configure Autoscale”
4. Configure:
   • Operation: deployment_autoscale
   • Deployment ID: steps.create_deployment.output.deployment_id
   • Min Replicas: params.min_replicas
   • Max Replicas: params.max_replicas
5. Click Save

Step 5: Add Conditional Success Check

Let’s verify the deployment succeeded:

1. Drag Conditional Branch after autoscale
2. Configure condition:
   steps.configure_autoscale.output.status == "active"
   
3. Click Save

Step 6: Set Pipeline Outputs

For the success path:

1. Drag Set Output connected to Conditional Branch (true)
2. Configure:
   • Key: “deployment_status”
   • Value: “success”
   • Key: “deployment_id”
   • Value: steps.create_deployment.output.deployment_id
3. Click Save

For the failure path:

1. Drag Fail action connected to Conditional Branch (false)
2. Configure:
   • Message: “Deployment failed - autoscaling not active”
3. Click Save

Step 7: Add Logging

Add visibility with log actions:

1. Drag Log after health check
2. Configure: “Cluster health verified ✓”
3. Repeat after each major step with appropriate messages

Your pipeline should now look like this:

Step 8: Save and Test

1. Click Save in the top-right
2. Click Execute
3. Enter test parameters:
   {
     "cluster_id": "cluster_abc123",
     "model_id": "model_xyz789",
     "deployment_name": "test-deployment",
     "min_replicas": 1,
     "max_replicas": 3
   }
   
4. Click Run

Step 9: Monitor Execution

1. Switch to Runs tab
2. Watch your pipeline execute in real-time
3. Click View to see detailed logs
4. Verify the output shows deployment status

Step 10: Set Up Triggers (Optional)

Automate execution:

Manual Trigger

Already configured - use Execute button

Scheduled Trigger

1. Go to Triggers tab
2. Click Add Schedule
3. Enter cron: 0 2 * * * (daily at 2 AM)
4. Click Save

Event Trigger

1. Click Add Event Trigger
2. Select event: model.updated
3. Add filter: model_id == params.model_id
4. Click Save

Now your pipeline runs automatically when models are updated!

Best Practices Applied

Health Check First: Always verify infrastructure before deployment

Parameterized: Reusable across different models and clusters

Error Handling: Conditional branching with explicit failure

Observability: Log actions at key points

Clear Outputs: Structured outputs for downstream systems

Troubleshooting

Health check fails

Cause: Cluster not ready or ID incorrect

Solution: Verify cluster status in Clusters page, check cluster_id parameter

Deployment creation fails

Cause: Invalid model ID or insufficient cluster resources

Solution: Verify model exists, check cluster capacity

Autoscaling not active

Cause: Deployment didn’t start successfully

Solution: Check deployment logs, verify resource requests

Next Steps

Pipeline Concepts

Understand DAGs and actions

Quick Start

Build your first pipeline in 5 minutes

Troubleshooting

Common issues and solutions