Is it possible to use Langchain or have an LLM read my data in the Lakehouse and generate results?

[supri03]

Hey guys !

I would like to work on having an LLM read my data in the Lakehouse (It's in the Tables folder) and then as I prompt it , it gives me an answer. Is this possible ? If yes, then please guide me. 

If there are any prerequesites like Azure Access etc, please let me know in detail. Thanks

Thanks a lot !

[vinchit]

Thanks for posting this query @supri03 . Even I am working on one such SQL LLM use-case where currently I am using SQL connection string via MS SQL Server and interacting with it via LangChain. But I want to also know if we can directly do with DB/Warehouse in One Lake?

[Anonymous]

Hi @supri03 
Thanks for using Fabric Community.
At this time, we are reaching out to the internal team to get some help on this. We will update you once we hear back from them.
Thanks

[vinchit]

Looking forward to your response @Anonymous 

[supri03]

Sure, Thanks for the response. 

[Anonymous]

Hi @supri03 
Fabric lakehouses are accessible(given proper permissions) just like ADLS Gen2 over ABFSS protocol, and data in the lakehouse can be used to finetune a LLM.

Hope this helps. Please let me know if you have any further queries.

[supri03]

oh ok, can you please guide on how to do this? 

[Anonymous]

Hi @supri03 
The internal team is working on this and I will let you know when there is an update.
Thanks.

[Anonymous]

Hi @supri03 
The internal team has confirmed that there are planning guidance on this in form of a blog. So once the blog is out, I will update the link here.
Appreciate your patience.

Thanks

[vinchit]

HI @Anonymous ,

Any updates regarding the blog?

[Anonymous]

Hi @vinchit 
Once I get the update from the team, I will post it here.
Appreciate your patience.

[Anonymous]

Hi @supri03 
We haven’t heard from you on the last response and was just checking back to see if you got some insights regarding the query. Otherwise, will respond back with the more details and we will try to help.
Thanks