OpenAI Timeout - Need to revisit

[bandelero]

Hey Pabbly folks,

I know there's an error handler on the way, which will, hopefully, help with OpenAI erroring out due to model overload.

Yet, very often, OpenAI works well, but it's just slow. And if you need to produce a large piece of text (which you cannot chunk into separate requests), the 40-sec limit is not enough.

I know OpenAI currently doesn't offer a way to handle requests in an async way (e.g. Pabbly makes a call to OpenAI, gets an execution id, and then pings a webhook when it's done). And they may never get to it given the popularity of their services.

Yet, we need to be able to use OpenAI with Pabbly. Hence, I am asking Pabbly team to revisit the timeout for OpenAI modules and make the equal to what Make has - e.g. 300 seconds.

E

 

[Supreme]

Hey @bandelero

Thank you for bringing up your concerns regarding OpenAI's performance and the limitations it poses. We understand that while OpenAI generally functions well, it can be slow at times, particularly when generating large pieces of text that cannot be split into separate requests within the 40-second time limit.

We acknowledge that OpenAI currently lacks support for handling requests in an asynchronous manner, such as using a webhook to notify when the execution is complete. We also understand the importance of being able to use OpenAI with Pabbly.

While we cannot guarantee the feasibility of revisiting the timeout for OpenAI modules and matching it to the 300-second limit of Make, we appreciate your feedback and will certainly take it into consideration.

 

[kotakeio]

Where are we at on this? I have workflows that won't work because of OpenAI modules timing out.

 

[Supreme]

Where are we at on this? I have workflows that won't work because of OpenAI modules timing out.

Could you kindly provide the task history ID associated with the issue you are currently encountering?