Call API Timeout, Webhook Timeout, and Open AI User Aborted Request

I’m wondering if there’s any way to get around time outs with the call api action and user aborted request errors with the open ai integration?

I’m getting the user aborted request error frequently from the chat completion open ai integration (it takes gpt a while to complete the intended task, usually > 60 seconds).

When I try to use the Call API action instead, I run into the same problem. Any idea if an how to work around this? I might try to use a webhook but would prefer to not use 2 updates

@NoCodeAndy I wanted to bump this. Is there anything we can do about timeouts for Open AI, Call API and Webhooks? I’m getting timeout errors on all of the aforementioned when requests exceeds 60 seconds. This is limiting, particularly when AI takes a bit to perform intended function.

Thanks for tagging me. Let me check with the team. :slight_smile:

1 Like

@NoCodeAndy one thing I’ll add to this, seems to be related though maybe not directly is that responses from Open AI using the Open AI integration seem to get cut off after a certain amount of time (less than 60 seconds). Was testing earlier and getting almost entirely partial responses when using the integration. Identical prompt used with same model, settings, etc. via an API call outside of Glide produced the full response.

Can I get your app ID?

I’ll pass that along to the team, it’ll help with troubleshooting.

Your Team ID is in the URL of your Glide dashboard, e.g.
go.glideapps.com/o/your-team-id-here

Your App ID is in the URL of the Glide builder, e.g.
go.glideapps.com/a/your-app-id-here/layout

Hi @NoCodeAndy , just wanted to raise my hand to ask about this as well. It’s the number one error I get in my app (which is an AI copywriting app, so you can understand the importance). Other companies have gotten around this by streaming responses in the API calls - would love to see that (or any) fix to this.

1 Like

@NoCodeAndy is there any update on this?

@J_Woosley @bru - How big is the payload you’re sending over? How many characters, approximately? I’ll pass this along.

The max for the OpenAI 4 Turbo model is 100,000 characters, but we are way under that. It shouldn’t matter, I get time out errors all day long. Would love to see streaming implemented!

@NoCodeAndy The payload out of Glide is ~ 6,500 characters and the response is ~12,000. Does that factor in to the timeout?

As a nearer term solution, I’d love to simply stop getting the webhook timeout error notifications. There is a call api function involved with the webhook that ultimately gets the data I need in Glide despite it timing out on the glide side, so I don’t want users to think there is an error when the action still effectively works. Removing the timeout or giving users ability to set timeout value across the board (Call API, Webhook, Open AI integration) would still be much more preferable.

@J_Woosley @bru - Thank you both. :pray:

Will update this thread when I have more information from the team.

I’ve escalated this in the meantime.

@NoCodeAndy Is there any update on this? Thank you very much!

Our engineering team is investigating the timeout issues now.

I’m keeping an eye on that ticket, along with this thread.

When there’s more to share I’ll drop it here. :slight_smile:

2 Likes