Add function calling params in OpenAI config

[baptisteArno]

No description provided.

[vforvilela]

Hey there! I'm in desperate need of this. :D
Right now, I'm cloning the OpenAI block and directly embedding my functions into it to bypass any interface complications.

Is there anything I can do to facilitate the advancement?

[Vinny-growth]

"I'm also facing the same issue, urgently needing this."

[baptisteArno]

Can you provide a specific use case for it?

[floux-digital]

Baptiste, thanks for the great work!

We are working to integrate assistant functions into the bot flow, but we are faced with a lack of this functionality. Currently we can define a function in the Assistant settings on the OpenAI platform, but the response, which must be JSON, is not recognized by typebot and so the conversation flow is interrupted.

We believe there are several ways to handle this, including a proxy using a lick function for the entire conversation flow, but that way we would only be solving our problem and not collaborating with the community.

Therefore, we thought of some hypotheses for this functionality in typebot:

• The simplest: maintain the function configuration on the OpenAI platform and in the integration settings, when an Assistant is selected, in addition to the treatment for Messages and Thread, also include the functions. Thus, the bot manager could create a step to test whether the response is a function, and if so, run a routine before returning a response to the user.

• The intermediary: create a middleware block called OpenAI Assistant Functions, which directly evaluates the response to identify whether it is a function. In the settings of this block, the user could define the functions and direct the routines to each of them.

The use case examples are endless, given that the structured responses received can serve as a payload for any type of application, from the simplest ones like creating a To-Do List or filling out a CRM to the most complex ones like managing technology stacks.