Using SK to build a chatbot on data in Databricks

[ashit-singh]

I've a requirement to build a chatbot on some tables residing in Databricks Unity Catalog. I was thinking to do it the old school way, a stateless Q&A (NLtoSQL) chatbot without any conversation history. But then stumbled across Semantic Kernel and the MultiAgentic approach.

I do have lots of doubts though, and Google search hasn't helped.

1. Is there a Databricks Unity Catalog Plugin I can use out-of-box?
2. Can I define a Databricks UC plugin and use that? Is it going to be supported?
3. Do I even need a Databricks UC plugin, if all my code is going to reside in Databricks?
4. I want answer to questions which will require some kind of similarity matching on the data stored in some of the tables, to find the top matches, and then do aggregation operations, etc.. Will writing a function to do this and adding it to the Kernel work as expected? Can I get some relevant blog or sample code for this?

Apologies if these questions seem naive, I've literally started reading about this last night.
And thanks in advance for the help!

[sophialagerkranspandey]

We recommend using the AI Foundry connector to Databricks for your needs.

[ashit-singh]

Thanks for answering.
Follow up questions:

1. Do I even need a Databricks UC plugin or connector, if all my code is going to reside in Databricks?
2. Where can I find some sample code for building a multiagent system in Databricks using semantic kernel?

Thanks again!

[sophialagerkranspandey]

For question 2, in Semantic Kernel, we don't have a sample of this.