Hiya! First time here. Was really liking openhands...
# openhands-cloud
l
Hiya! First time here. Was really liking openhands while testing but i'm suddenly getting this error when i'm trying to test the "openhands" label in github. Is there anyway to get better error logs? Also as a side question, is there a way to influence the behaviour of this openhands label skill? Would using an explicit SKILL.md file in my repo or just a top level AGENT.md file work best? Thanks in advance!
oh waveq 1
m
Do you have a profile active? Or defaulted? Because i can see in the img that it is written select a model. For that reason nothing is happening i think
l
Oh weird, ill try again there again But just as an aside, i also did try to run openhands locally and im getting this error here too
Unexpected completed event: <class 'litellm.types.llms.openai.ResponseFailedEvent'>
okay again, it seems like it's handing and not providing any reponse again and i do have a model selected. Here's details on the openAI model im using
m
Ok let me try to reproduce. Are you suing openhands local GUI or coud?
l
Cheers thank you, im just trying to use it in Openhands cloud Model used: OpenAIopenai/gpt-5-mini-2025-08-07 and i tried to get it to look at an issue when labeled as openhands, but im also trying to talk to it normally and am not getting a response and it's just that generic error Tried in firefox w/ ad blocker and google chrome w/ no ad blocker
m
Could you try with gpt-5.5? Just if the convo is starting. I suppose there is a problem with this model
l
same thing sadly. I'm looking on openAI and it's saying that there were some issues, but not sure if it's related
i can try out some other models tomorrow though worse case scenario to see if the problem persists
c
This should have been fixed in agent-sdk 1.32.0.
🎉 1
It happens only for GPT models, FWIW, and I’m not sure if it’s deployed or still bugged
l
okay phew that's great to hear! ill probably just give it a day or two. what was the issue causing it if you dont mind me asking?
c
The cloud uses Laminar for observability, and we enabled streaming for LLM responses: the current Laminar doesn’t play nice with the Litellm types we were using for streaming chunks and responses
🙏 1
We didn’t see it, because Laminar is not normally enabled (not on our dev machines usually)
Personally I met it too, on the cloud, and for a day or so, I switched to a Claude… though maybe I should use the time to test other LLMs!
l
looks like it's still bugged as of this morning. Just confirming, did you say that Claude and other models are working fine for now?
c
yes
✅ 1
l
Heyo, just popping a message here to say that GPT 4.x is working, but looks like this issue is still popping up for GTP 5.x models