I need to restart Cursor frequently because I'm constantly getting stuck in "Generating...."

Support Bug Reports
21

image
image1669×1205 456 KB

image
image1698×1214 345 KB

Full console log

I asked an agent (Ctrl + I) to write code, then asked to fix because it wrote the wrong file. Then did some adjustments to this and other files (used cursor tab).
Then selected few lines of code and asked a question about it (Ctrl + L) and it stuck.

UPD

Few minutes later it unstuck with a connection error. The next request was successful, so not sure it was an issue of the topic.

image
image1169×314 24.5 KB

UPD 2

I guess it finally stuck. There are no errors in console when request is “running”, until I press the Stop button.

image
image1920×1345 245 KB

Full log 2

Version: 0.46.2
VSCode Version: 1.96.2
Commit: 84b9c6d907219bb8c2874f299540eb6a079187a0
Date: 2025-02-23T01:17:13.048Z
Electron: 32.2.6
Chromium: 128.0.6613.186
Node.js: 20.18.1
V8: 12.8.374.38-electron.0
OS: Linux x64 6.8.0-52-generic
22

First, I always say “continue” when it gets stuck generating.

Second, if the the conversation is requiring a long response, this hass really helped:

Last night I was deep in context and I told it to do something and it was taking FOREVER – so I said “STOP”

Then I said

"Break the response down into smaller steps, and give me a reasonable chunk at a time, to avoid too long a context window"

This worked BEAUTIFULLY and it broke the response into SEVEN reply chunks, that I then stitched together.

Finally, Ive found 45.14 to be the most stable Agent version for me.

I updated to whatever 46.8 – and it was completely unusable, so I went back to 45.14

image
image394×195 4.22 KB

23

Yes but you are using 0.45.14 which means you do not experience 0.46.x related issues so its a different topic. on 0.46 yes this is the case, we have to restart often

24

I am experiencing this issue as well and it’s counting the stuck generations against my monthly limit, which is pretty annoying since it takes multiple tries to get it to complete.

Version: 0.46.8 (Universal)
VSCode Version: 1.96.2
Commit: be4f0962469499f009005e66867c8402202ff0b0
Date: 2025-02-28T19:32:53.996Z
Electron: 32.2.6
Chromium: 128.0.6613.186
Node.js: 20.18.1
V8: 12.8.374.38-electron.0
OS: Darwin arm64 23.6.0

25

Yeap… Same situation here… just downloaded the application, subscribed to pro and all the time is the same, works for 2-3 chats and stop to works !! :frowning:

26

PS: I have just figured out that there is no support link/email/form on the websiste. !! :frowning:

27

I’m on mobile but please look at my post just…

Tell the bot to give the response in smaller chunks to avoid context limits.

28

Same issue here

29

This issues have NOTHING to do with context. If the IDE does not send requests to AI, has aborted tool usages etc. its not about context.

30

well, what I have found with respect to “context” is that its a black_box as far as we dont truly know the depth of context - especially when talking to whatever through the abstraction of Cursor infra – thus, what I have empirically found is that telling the bot to chop a long reply into smaller chunks that do not hit the context limit works REALLY well, insomuch that the BOT knows its context length, and responds accordingly.

Give it a try.

WRT “not getting the request” - thats why I said to flap your login to Cursor. As I found this also helps refresh the pipeline…

Also, this thread is great:

31

It makes no sense to shorten a 1 line request that requires a 1 line response from AI. if the network request is not made to servers there is nothing to process by AI. the context is irrelevant in this case.

if i send a prompt with content “Test” the AI usually responds with “I see you are trying to test the response, please let me know how i can help…”. there is nothing to split in this context.

32

I cant help in that context…

–

What version are you using?

Also try the logout and login thing I mentioned and Report Wakaowski!

33

That always happened to me. I’ve being using Cursor for 4 months now.

34

Well I’ve just cancelled my subscription so I can try using the W competitor as soon as my subscription period expires in 3 days.
I heard is more stable and a pretty good alternative.
I’ll let you know how it goes :slight_smile:

PD: It’s kind if childish this forum doesn’t allow to mention the competitor’s name just as if it was Vold… :stuck_out_tongue:

36

Same issue !!!

37

Not sure if my issue is the same but sounds similar. Cursor for me yesterday and today has been “not responding” and even though I’ve restarted it and restarted by machine, it continues to do this and keeps giving me an AI chat error “apply model” is broken or not working, it says “edit tool” is not working and broken, and it won’t edit any files for me. When I do the toggle console log, I get a bunch of errors in console log for the application. At this time it’s unusable to me. I’m on the latest version 46.■■.

38

Same issue

39

I’ve started my 14 days of pro free trial of W and I can confirm the “generating…” problems and constant request failures are all related directly to Cursor. I´ve used the same 3.7 sonnet model there and it worked without a hitch. So Cursor needs to get its :poop: together an act to fix its things.

40

Same here. Claude-3.7 is having these generating problems

image
image1544×348 35.8 KB

41

same issue here, VERY difficult to work with. I’d rather avoid rolling back to a previous version or try a competitor but if this is not fixed soon i’ll have no choice.