It’s pretty sick. Just got to remember that they limit you to a 100 requests per day while it’s in preview
gpt-3.5-turbo-1106 is lightning fast…early testing shows no errors but will continue to test and see how we get on.
I spoke too soon.
Now im getting “Bad Gateway” & “Service Unavailable”
Unknown error: {
“status”: “come back later”
}
Getting more random now
Turns out its an outage 
3 Likes
I’m getting this with the OpenAI speech to text action. They are long audio files of about 30mins (but highly compressed so within the file size limit for OpenAI). I’m getting charged on my OpenAI dashboard but nothing is returning to the selected column in Glide.
It feels like Glide is timing-out.
If you don’t mind using Make, you can use Deepgram. It’s much faster.
I couldn’t find a deepgram module, but thanks for the reference, Deepgram looks interesting.
But staying on topic, I would like to know what’s causing this.
I have no insights, but the likeliest outcome in my opinion is the call times out before a response can be received.
DeepGram does not have a module in Make. You would have to use a HTTP module. Please let me know if you need any assistance with that.
Yes I use HTTP module quite a lot, although now we’re a far reach from the feature I expected to work in Glide.
Thanks for offering.
1 Like
Just curious any update on this? I still see the error on my end. Is the solution to use Make/Zapier?
Yep , still getting these for a lot of my OpenAI columns
Have you tried the Glide-Ai columns? The generate text and text to choice were just updated as well.
No , I have many many OpenAI columns so dont really want to have to switch them all
I’m still encountering the same issue, but now it’s behaving quite oddly. My setup involves multiple “Complete chat” actions within a single action. Interestingly, it functions correctly for the first three instances, but I encounter an error on the fourth or fifth action, which also disrupts the logic of my app. Additionally, I would appreciate more advanced debugging tools for situations like this, such as a “wait for condition” feature that not only notifies but also triggers an action in case of failure or completion.
Have you enabled logging? Would be interesting to see if there are more details in the log.
Does anyone have a solution for this issue? I am also encountering the error message “The user aborted a request” when transmitting large tokens to OpenAI. According to the documentation, I have not exceeded the tier limits. The process functions correctly when executed independently; however, when multiple triggers are chained within a workflow, the error “The user aborted a request” occurs.
How big is the input you’re sending over? I think there’s a chance OpenAI takes too much time sending a request back that Glide “aborted” the request due to a timeout.
Each trigger has 25000-50000 tokens. I have 23 of them in sequence so more than 1,000,000 tokens. Gemini was able to process this but Open AI has issues.
1 Like