A ModelContextProtocol Server(mcp) for Scoring with SAS Viya - Part 2: Persistence
Views
536
Introduction:
In the previous article I introduced a mcp server for scoring with SAS Viya. The next logical step is to persist scoring results for post processing. The old me would have added added some code to the scoring tool to persist the data. But with mcp servers I can pick and choose how and where I want to persist the data with little or no coding. See this link for some mcp servers for data bases.
In this article, I am using the mcp/redis server to publish the scoring results to a SAS Viya CAS table. This allows me to post process the results immediately with SAS Viya. The picture below shows the flow. Given the architecture, I can replace the entire persistence flow with different set of mcp servers.
Why Redis? A few years ago, I worked on a POC which required streaming the scores (using SCR) via Redis. The solution consisted of a small set of custom docker containers in a pipeline.
Key points:
.
- The boxes in blue are the custom code I developed for this article
- The viyascoring mcp is the custom mcp server that was described in the previous article.
- mcp/redis is an open-source server from Redis. It requires a Redis server
- The Redis server is the non-commercial version of Redis. This is ok for the demo, but you know the drill - for production systems get a commercial license.
- The subscriber is custom code
- Subscribes to a specific channel( "score" in this demo)
- The received message is used to update a master table in SAS Viya
- The mcp server code is available on github at this link
- The redis subscriber code is available at this link
Solution in Action
Finally
Hopefully I have convinced my fellow SAS developers to enable their users to access the models using prompts.
If you do not know JavaScript or Python partner with an expert in those languages.
