Please upvote auto re-execute failed tasks (if you use OpenAI, you know you need this)
Status: Open · Asked by Evgeny Bik on · 0 views
Folks,
If you're using OpenAI, you know your modules will be erroring out from time to time due to OpenAI models being overloaded. Yet, as my experience with Make/Integromat showed, if you simply re-query the same OpenAI endpoint after it failed, it works and produces output.
Therefore, we really need automatic retry capabilities with Pabbly for failed modules. Otherwise, we'll always have to manually re-drive the modules (which is suboptimal CX).
Please upvote this feature here:
https://pabbly.hellonext.co/p/re-execute-failed-tasks-automatically
Cheers,
Eugene
Hi @bandelero,
We are very much aware of this issue, and rest assured, our technical team is working to find an optimal solution for it. In the meantime, I would suggest checking out this post on our Facebook Group where one of our users has implemented a workaround to avoid timeouts from OpenAI.
https://www.facebook.com/groups/pabbly.connect/permalink/1388804011899217/?mibextid=Nif5oz
Hi @bandelero,We are very much aware of this issue, and rest assured, our technical team is working to find an optimal solution for it. In the meantime, I would suggest checking out this post on our Facebook Group where one of our users has implemented a workaround to avoid timeouts from OpenAI.
https://www.facebook.com/groups/pabbly.connect/permalink/1388804011899217/?mibextid=Nif5oz
Thank you!
That workaround seems to solve for errors due to timeout. However, often OpenAI errors out due to model overload. But re-querying the same endpoint right after the error works most of the time.
Die this workaround solve for auto retries?
Eugene
That workaround seems to solve for errors due to timeout. However, often OpenAI errors out due to model overload. But re-querying the same endpoint right after the error works most of the time.Die this workaround solve for auto retries?
The workaround is only for errors due to timeout. For errors caused by other factors, we will introduce a more direct solution.
Thank you. Given the urgency of this feature, is there an ETA?
Adding a new feature across the platform requires a lot of testing before it's implemented. For now, I won't be able to give you an ETA.
The problem is there is also not enough tracking within Pabbly to understand which rows in e.g. a Google Sheet has been executed and which one not.
There should be a list with all requests recevied and the status whether executed or not. Then there should be a button to push re-execute or a setting to auto-execute again.
There should be a list with all requests recevied and the status whether executed or not.
You can find these details from your Task History.
Then there should be a button to push re-execute or a setting to auto-execute again.
In your Task History, you do get an option to manually re-execute tasks one by one or in bulk.
youtube.com/watch?v=GOJt-reqvJA
Also, we have introduced the Auto Re-execution feature, which you can use to automatically re-execute failed tasks.
https://forum.pabbly.com/threads/how-to-enable-auto-re-execution-for-workflows-coming-soon.15088/