To close out 2025, SAS Model Manager added two new publishing destinations in the 2025.12 release: SingleStore through SAS SpeedyStore and SAS Event Stream Processing (ESP). SAS SpeedyStore is based on SingleStore, which is a database that delivers elastic scalability, high availability, and exceptional price performance. As a publishing destination for SAS Model Manager, users can publish SAS models to a SpeedyStore SingleStore destination for performant scoring of SAS models in batch. SAS Event Stream Processing is a solution on SAS Viya for streaming analytics and IoT use cases. As a new publishing destination for SAS Model Manager, users can publish a SAS model into an ESP container for use within SAS Event Stream Processing. Let’s dive into each destination.
SpeedyStore SingleStore
Before you can publish a model to SingleStore, you’ll need a SingleStore instance. Additionally, the SingleStore publishing destination uses SAS In-Database Technologies, which is included in the SAS Viya Enterprise package or the SAS Speedystore offering (which both also includes SAS Model Manager) or can be licensed on its own. With all of your technologies in place, an admin will need to configure SingleStore for PROC ACCELERATOR.
To create the SingleStore publishing destination for SAS Model Manager, an administrator will need the SAS Viya Command Line Interface (CLI). To get started, admins can download the SAS Viya CLI, create a profile, sign in, and install the models plug in. With their CLI set up, the administrator can now create a SingleStore Publishing Destination. To run the CLI command, your admin will need to know the name and port number of the database they would like to publish their models and decide who should have access to the destination.
sas-models-cli destination createSingleStore
--name nameForTheDesintation
--credDomainID nameForTheDomain
--identityId userOrGroupIDThatHasAccess
--identityType selectUserOrGroup
--databaseHost nameOfDatabase
--databasePort portNumberoOfDatabase
--databaseSchema model
--singleStoreUserId userIDforSingleStore
--singleStorePasswd password DforSingleStore
From SAS Model Manager, users can now publish their SAS Data Step, DS2 embedded process, and DS2 multi-type models to the SingleStore destination in just a few clicks. Once published, users can score data using their deployed model using publishing validation or through SAS code using the RUNMODEL statement in PROC ACCELERATOR. To see this new publishing destination in action, check out the following demo video:
SAS Event Stream Processing
To create a publishing destination for SAS Event Stream Processing (ESP), organizations must first license SAS Event Stream Processing. Both SAS Model Manager and SAS Event Stream Processing are available in the SAS Viya Enterprise package, but both tools can be added to other packages as well.
Before you can create the publishing destination in SAS Model Manager you must add an container registry in SAS Event Stream Manager. From SAS Event Stream Manager, go to settings, then select Container Registries. Add a container registry here, but remember the name, because we will need it for CLI. Just as we did for the SingleStore destination, users can also leverage the SAS Viya Command Line Interface (CLI) to create the ESP destination.
sas-models-cli destination createESP
--name nameForTheDesintation
--targetContainerName nameOfContainerInESP
Users can publish their SAS models into an ESP container available for use within SAS Event Stream Processing. We cannot run publishing validation for the publishing model in SAS Model Manager, but the model can be picked up and used within ESP flows. Check out this demo to see an example of model publishing to SAS Event Stream Processing:
Conclusion
SingleStore is another destination built on PROC ACCELERATOR, just like Databricks. We are planning to expand our integration of PROC ACELERATOR in SAS Model Manager by adding our Cloudera and Synapse as our next publishing destinations, but what publishing destinations would you like to see next? Let us know in the comments!
