SDA (Structured Data Analysis) Plugin Level Toggle

Hi ​@nicholas.marr – thanks for your feedback.

That’s a really interesting note. We designed SDA to trigger whenever the reasoning engine believes that it will do a poor job without it. It might be helpful to review a few examples where it’s overtriggering so we can incorporate your feedback into our Agentic Reasoning Engine.

Do you mind sharing more details about the cases where it’s not quite behaving as desired? ​cc: @anayak 

​@Ajay Merchia 

Examples i saw it on included getting ticket details from ServiceNow work notes (we are getting tracking numbers) and Splunk report data (network traffic and server health checks). We still need SDA for certain things and even when it misfires it’s not terrible. Main reasons I'd like the toggle include:

• SDA adds 5-10 seconds of time to the plugin i noticed
• SDA ignores format instructions and uses it’s own styling

Hello,  I currently have something built where SDA is directly obstructing my intended use case. My use case allows managers to pull tickets for all of their direct reports in ServiceNow, as well as tickets for a specific direct report. I’ve set sysparm_limit to 200 to avoid a timeout. For example, if a manager has 60 direct reports with 5 tickets each, that’s 300 tickets total, but the results are capped at 200 by sysparm_limit.  If the manager first pulls tickets for all direct reports, then tries to use my other plugin to pull tickets for a specific direct report whose last name starts with “W”, data analysis runs instead of my direct-report plugin. As a result, SDA reports that this direct report has no tickets, because it does not run either plugin a second time. Below is an example. The LLM directly asked me to request tickets for a specific direct report such as the employee mentioned. When I attempted to search for that employee, and after running SDA, it told me there were no tickets in that data set for that employee.