Wallaroo

Send event data to Wallaroo for real-time ML model inference, pipeline execution, and AI-powered event enrichment.

⚠️

This integration is currently in beta. Configuration and behaviour may change.

Overview

The Wallaroo integration delivers events server-side from Datafly Signal to a Wallaroo ML pipeline for real-time inference. This enables use cases such as predictive scoring, fraud detection, recommendation engines, and dynamic personalisation — all powered by your deployed ML models without adding client-side latency.

Prerequisites

Set Up a Wallaroo Instance

You need access to a Wallaroo instance (cloud or self-hosted). Sign up or log in at wallaroo.ai.

Deploy a Pipeline

In the Wallaroo dashboard, navigate to Pipelines.
Create and deploy a pipeline with your ML model(s).
Note the Pipeline ID from the pipeline detail page.
Ensure the pipeline is in a running state.

Obtain the Model Endpoint URL

On the pipeline detail page, find the Inference URL (the HTTP endpoint for your deployed pipeline).
Copy the full URL, including the path (e.g. https://my-instance.wallaroo.ai/v1/api/pipelines/infer).

Generate an API Key

In the Wallaroo dashboard, go to Settings > API Keys.
Create a new API key with inference permissions.
Copy the API key immediately — it will not be shown again.

Configuration

Field	Type	Required	Description
`api_key`	secret	Yes	Your Wallaroo API key for authenticating inference requests.
`pipeline_id`	string	Yes	The Wallaroo pipeline ID to send events to for inference.
`model_endpoint`	string	Yes	The full URL of the Wallaroo model inference endpoint.

Signal Setup

Management UI

Go to Integrations and open the Integration Library tab.
Find Wallaroo or filter by Other.
Click Install and fill in the required fields:
- API Key: Your Wallaroo API key
- Pipeline ID: The target pipeline ID
- Model Endpoint: The full inference endpoint URL
Click Install Integration.

Management API

curl -X POST http://localhost:8084/v1/admin/integration-catalog/wallaroo/install \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Wallaroo",
    "variant": "default",
    "config": {
      "api_key": "your-api-key",
      "pipeline_id": "pipeline-abc-123",
      "model_endpoint": "https://my-instance.wallaroo.ai/v1/api/pipelines/infer"
    },
    "delivery_mode": "server_side"
  }'

Delivery

Events are delivered server-side from your Datafly Signal infrastructure directly to the Wallaroo inference endpoint. No client-side scripts are loaded for this integration.

Signal sends event data as the input payload to your Wallaroo pipeline. The pipeline runs your deployed model(s) against the event data for real-time inference.

Common Use Cases

Use Case	Description
Predictive lead scoring	Score website visitors in real time based on behavioural signals
Fraud detection	Flag suspicious transactions before they are processed
Personalisation	Generate product recommendations or content suggestions per user
Propensity modelling	Predict likelihood of conversion, churn, or other outcomes
Anomaly detection	Identify unusual patterns in event streams

Testing

Enable the integration and trigger events on your site.
In the Signal Management UI, check the Integration Detail page for delivery status and any errors.
In the Wallaroo dashboard, check your pipeline’s Inference Logs to verify events are arriving and being processed.
Review inference results in Wallaroo to confirm your model is returning expected outputs for the event data.

Troubleshooting

Problem	Solution
401 Unauthorized	Your API key is invalid or expired. Generate a new key in the Wallaroo dashboard and update the integration config.
404 Not Found	The `model_endpoint` URL or `pipeline_id` is incorrect. Verify the endpoint URL and pipeline ID in the Wallaroo dashboard.
503 Service Unavailable	The Wallaroo pipeline may not be running. Check the pipeline status in the Wallaroo dashboard and ensure it is deployed and active.
Timeout errors	ML inference can be slow for complex models. Check your model’s average inference time in Wallaroo. If consistently slow, consider optimising the model or increasing Signal’s delivery timeout.
Inference errors	The event payload may not match the expected input schema for your model. Check your model’s input requirements and ensure the event data includes the necessary fields.
Events delivered but no inference results	Verify the pipeline has at least one active model step. An empty pipeline will accept requests but produce no output.

For full API documentation and pipeline setup guides, see the Wallaroo documentation.

Pipeline Router Overview