GLM - Timed out waiting for intercepted first chunk (45s)

[hydrange0]

Hello! I use IRP on a lower end device (older samsung laptop from like, 2014 or so, might be older) and I recently started experiencing issues with the GLM chat.
The chat will more often than not time out before I get a response in SillyTavern, but it will still send in the Z.AI chat right after.
This isn't an issue I had at all before I updated to the 2.6.1-patch, and I have to swipe multiple times until I can get an actual response in SillyTavern.

I'm running IRP on my laptop, while SillyTavern is on my phone (I use the LAN feature for the responses), if that could be the cause of any issues.
I'm not sure if this is something on the side of Z.AI itself, or if there are any changes you could make. If so, it'd be really appreciated if there was an option to increase the chat timeout time yourself, or something like that.

Once again, my bad if this isn't something under your control!! Thank you

[LyubomirT]

Hello! Do you mean that it fails to get the first chunk while the request is actually created? This sounds more like a server-side bug since Z.AI sends SSE (Server-Sent Events) chunks, IntenseRP doesn't manually request them, so receiving them depends entirely on whether the chunks are sent from the actual server. If they're not, this is almost definitely an issue on their end. But I'm also not denying that this could be something with IRP. Did it start happening right after you updated to 2.6.1, or did it not coincide that cleanly?

[LyubomirT]

It would also be helpful if you uploaded a logfile or a condump; they'd help me analyze the issue more deeply (since those are the actual system messages from IRP).

[hydrange0]

It wasn't that clean I think, I just know that it only started happening after I updated.
And I think so? I send in the request, wait a little bit, and then get the timeout issue. It says the request is completed and it sends an empty response in SillyTavern, while the actual response starts generating in Z.AI.

It also seems to coincide a lot with route fullfilling errors (route is already handled)

[REDACTED BY MAINTAINER]

[LyubomirT]

This is actually interesting, it seems like UI timeouts somehow got mixed together with API weirdness. I can't reproduce on my end, but the logs sure will be of great help. I'll give it a look now and see what I can find, though hopefully the issue is not as big as it may seem.

Also, just a side note, I've just redacted the condump in your comment since the URLs there contained your auth token (so that nobody steals your session :D).

[hydrange0]

Alright, thank you so much!! I hope you can find whatever is going on lol

[adhdandy]

I want to add I into the same issue using Deepseek, I would post logs but I had not turned logging on before.