I am trying to set up a Stream Analytics Job that accepts input from an Event Hub, processess the input via a ML model, and sends the output e.g. to a Power BI dashboard.
I deployed an ONNX model on an ACI (Azure Container Instance) instance following the documentation here: https://learn.microsoft.com/en-us/azure/machine-learning/how-to-deploy-and-where . This seems to be fine and I do get the automated swagger definition and can use the service via REST.
How can I connect to my ML deployment from within the Stream Analytics query? There is the "Functions" setting under "Job Topology" of the "Stream Analytics job" page, but I cannot figure out how to add it there. This ( https://learn.microsoft.com/en-us/azure/stream-analytics/stream-analytics-machine-learning-integration-tutorial ) suggests that it is possible, although it uses Azure Machine Learning Studio (as opposed to Azure Machine Learning without "Studio". I'm quite new to Azure and don't know if this matters or not but find it a bit confusing).
There is an ongoing limited preview which you can sign up to get access to this functionality. You will then be able to use the ONNX model you have deployed on ACI in your Stream Analytics job. We expect to roll out this functionality more broadly in the coming weeks :)