Awesome video! It is missing one piece of the puzzle from my video idea, although I see that it wasn’t completely clear from my comment: It would be to have your function to retrieve the data be a tool call (function) so that you can ask the assistant for any particular stock and it will execute your function to get that data. So the flow is: - Tell me about stock XYZ - Function call to retrieve JSON on stock XYZ - Upload the JSON data to the assistant thread - Let the code interpreter do its thing like you show What I struggle to succeed with is to have the output of your function call be a file. Normally, you would just get the JSON as a string and pass that back to the assistant via the “submit_tool_outputs” endpoint. But if you are dealing with a lot of data, that is going to be incredibly expensive very quickly. Therefore, it would be nice to simply be able to return a reference to an uploaded file in the “submit_tool_outputs” call. I hope that makes sense 👍
Okey I see :D that makes sence, will def take a look and change the code, I did try it with function calling and i did work, so should be no problem :)
As always "great work" Kris. Thx! I would have liked to see it get integrated in a GPT within the actions / schema. I find that very difficult to setup - did not manage to do it so far. It would be nice to see what it could do. Plus there wouldn´t be any API-costs - right?
HI, Your AI videos are just awesome. Can you pls make video of how to create csv of products of store with attributes like color, size and then fetch result by using openAI model according to user's input .
Awesome video! It is missing one piece of the puzzle from my video idea, although I see that it wasn’t completely clear from my comment: It would be to have your function to retrieve the data be a tool call (function) so that you can ask the assistant for any particular stock and it will execute your function to get that data.
So the flow is:
- Tell me about stock XYZ
- Function call to retrieve JSON on stock XYZ
- Upload the JSON data to the assistant thread
- Let the code interpreter do its thing like you show
What I struggle to succeed with is to have the output of your function call be a file. Normally, you would just get the JSON as a string and pass that back to the assistant via the “submit_tool_outputs” endpoint. But if you are dealing with a lot of data, that is going to be incredibly expensive very quickly. Therefore, it would be nice to simply be able to return a reference to an uploaded file in the “submit_tool_outputs” call.
I hope that makes sense 👍
Okey I see :D that makes sence, will def take a look and change the code, I did try it with function calling and i did work, so should be no problem :)
Very nice demo, thanks!
what bout using openai assistants api but on local llm using ollama???
I don't see how this does what was requested. It doesn't return a URL at all.
Like always its so nice thanks for sharing.
As always "great work" Kris. Thx!
I would have liked to see it get integrated in a GPT within the actions / schema. I find that very difficult to setup - did not manage to do it so far. It would be nice to see what it could do. Plus there wouldn´t be any API-costs - right?
noted :D
Is there way how to send data from openai chatt to third party service?
HI,
Your AI videos are just awesome.
Can you pls make video of how to create csv of products of store with attributes like color, size and then fetch result by using openAI model according to user's input .
Kris - is this project in github yet?
please provide code source
It's a cool feature but costly $0.03 / session