All-Hands-AI
Docker runtime slow through frontend only?
I am running OpenHands by building from source with the docker compose file. I have also tried the official image. Everything works, but the terminal commands take forever to execute, but what I find odd is that it only has a slow execution time from within the OpenHands interface.
If I execute commands directly within the docker runtime by attaching to the container, everything is very fast. Is there a hidden embedding model or something running inside the OpenHands interface. Why does it take over 2 minutes to return anything from npm i when executed from the OpenHands terminal, but execute super fast in the docker environment terminal?
I am not sure if this is a bug or just me, because it is just really slow when the using things like packages and for install packages. I have checked RAM availability as well, and running on M4 Pro in docker with the default runtime.
Seems its the problem of 0.19 version. All commands fail after default 2 minute timeout instead of returning the output right after finishing. There is no such problem in 0.18
There was an issue in version 0.19 which was fixed in 0.20. However, this may still be an issue. I've brought it up with the team to see if someone can take a look.
Thanks for the help. I started on 0.19 when I first downloaded last week. Tried 0.20 too, I'll try 0.18 now.
I can confirm it's going on in 0.20.0 It's actually really frustrating. Things are taking a long time. In 0.19.0 I was about to give a response which seemed to break the stall but in 0.20.0 that doesn't help. It's not even timing out because I don't think the process is even running any more.
Hey all I think we have some understanding on this now. For now you can use 0.18 version as that doesn't seem to have the issue. Hoping to get a fix soon 🤞
There's been some improvements in this area in 0.21. However, don't believe it's 100% fixed. Let me know if you all are seeing better/worse/same results if you happen to see this.
@mamoodi I am seeing better results, especially at the beginning of a conversation. The farther into the conversation I get the slower everything is running. It's kind of weird because in 0.20 the functions would get stuck and timeout but the conversation could go really long before everything would slow down. In 0.21 it seems like functions aren't getting stuck as much at the beginning but as soon as about 100 steps in everything slows down and lots of functions take a long time to complete.
I am also seeing better results, still using v0.21.1. And also seeing the eventual slow down, but its at least not being slow from the first request anymore.
Is this still happening? Things have gotten better I believe. There's a few issues when the agent gets into a bad state sending Ctrl + C or empty commands and those are being addressed.
New to the project, and found this issue because 0.28 terminal is very slow
Hi all, we might need more information that "very slow" to be able to debug the problem. I'm going to close this as completed for now, but if you're encountering issues with slowness in the terminal, please created a new issue with more information such as what prompt you ran, what version you're using, your operating system information, etc. You can link back to this issue as well if you'd like.
