Very valid point. This is basic example for everyone to understand on how it can be integrated, but u can add another AI API call in the loop to convert NL2SQL
Sometimes it’s not as simple as running a sql. User can ask anything which involves many unrelated tables. In that case it’s better and easier to use function calling. But instead of defining specific function, I just give a function that can run any query.
The way I made to talk mySQL DB was to ask LLM to extract data points from the query and construct a JSON payload with all require fields and then you can use it in an API call or in a SELECT query. Already did this with my conversational AI chatbot
why not ask the LLM to construct a SQL directly base on user enquiry and execute it, it will be much more impressive. As long as LLM read the database structure, it's achievable
Great video again, and first person on this video to give you a "like"
Thank you :)
Thanks for the video. A very newbie question: can the direct integration be done in CHATGPT or similar AIs that would not require programming?
I know you've seen the newcomer's comment, "How come I haven't seen your channel before"... but OMG...How come I haven't seen your channel before?
could be good but on prem. possible without API cohere?
What’s the point? Why not just run the sql and use the output directly if you’re already writing sql
Exactly my thoughts... Was hoping to finally get some sort of NL2SQL but this is just a basic function call.
Very valid point. This is basic example for everyone to understand on how it can be integrated, but u can add another AI API call in the loop to convert NL2SQL
Sometimes it’s not as simple as running a sql. User can ask anything which involves many unrelated tables. In that case it’s better and easier to use function calling. But instead of defining specific function, I just give a function that can run any query.
So that your employer can pay you more for wasting time 😂
The way I made to talk mySQL DB was to ask LLM to extract data points from the query and construct a JSON payload with all require fields and then you can use it in an API call or in a SELECT query. Already did this with my conversational AI chatbot
Nice. Will try to create one similar
Nice Mervin as always.. quick qn.. have you done one with a sql agent?
Thanks a lot, it would be awesome if you can integrate this (DB integration) into an autonomous agent case (i.e crewai, autogen etc)
Yes good idea
@@MervinPraison awesome :-), looking forward to it.
can we do this with opensource model from hugging face
This is great and is useful for many applications like chatbots.
Yes
Can you talk fullstack app but on custom GPT
Great!
Thanks
why not ask the LLM to construct a SQL directly base on user enquiry and execute it, it will be much more impressive. As long as LLM read the database structure, it's achievable
Yes it is a nice idea. Will create a video to include that . This one I was focusing mainly on DB integration for beginners
Allowing an AI model to directly create and run queries is insecure and dumb.
thanks!
Unecessary costs for something that can be done also with python etc