ntjohnson1 commented on issue #1394:
URL: 
https://github.com/apache/datafusion-python/issues/1394#issuecomment-4142920048

   One other point here for visibility. I am personally a big fan of dataframe 
APIs. However, as we've been experimenting with agents writing more datafusion 
using the SQL parser shows way more immediate success just based on the volume 
of SQL already in the training datasets. This probably isn't datafusion 
specific.
   
   We have a couple minimal skills for pointing LLMs to explore the schema 
first (since many of the errors are just related to not understanding schemas) 
then have pretty good luck with one shotting the SQL based on "can you write a 
query so I can see <topic or quantity of interest>". So far we haven't gone 
super deep into optimizing queries with agents when they start being 
slow/hitting large amounts of data.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to