Hello! :sunny: Some feedback on 1st setup experien...
# proj-agent-canvas
c
Hello! ā˜€ļø Some feedback on 1st setup experience: I used the Docker route to setup OpenHands and I was able to bring up easily the backend + frontend, but when I first opened the frontend (127.0.0.1:8000) I was seeing errors in the top right corner (401). After a few online searches, I figured it would be because some configuration was missing. I tried following some documentation about setting your LLM (which I couldn't in the first splash screen because of the errors). I couldn't follow the documentation because some portion of the UI wasn't showing. I did find the message "SDK settings schema unavailable.". It's very unclear that this "message" is an actual error message, and the UI that I'm seeing is actually not the expected UI. Considering I was seeing errors on the top right, I wasn't expecting this message to be an error itself. I couldn't find anything on the documentation (web) about this, if I search this string in the Github repo, nothing relevant pops-up. I was wondering around until I found this md doc: https://github.com/OpenHands/agent-canvas/blob/rel-1.0.0/docs/SELF_HOSTING.md which contained a lot of really great info which can't be found online (web). Reading this, I was able to figure out that the problem was that the front end needed the API key to be set in the "backend" config. 1- I would have found this more easily if the backend wasn't "half" configured already 2- Online documentation should either contains some of the documentation from Github, or you should redirect people to the Github documents (even in the "Quick start" doc). This doc was also very appreciated to get started: https://github.com/OpenHands/agent-canvas/tree/rel-1.0.0 3- Too many repeating documents is very confusing. I'm not done setting everything up yet, but I hope this "first xp" will be helpful to someone else! Cheers!
šŸ™Œ 4
ā¤ļø 3
b
Thanks for that feedback.
h
Hey! Thank you for the breakdown on the hurdles you are ran in to! I have a couple of questions if you get a chance. 1. Was the Docker install on a VM or on your local machine? 2. Are you running macOS, Linux or Windows? 3. Did you start with the install steps here https://github.com/OpenHands/agent-canvas#option-2-with-a-docker-sandbox or somewhere else? If you would prefer to talk through I have a scheduling link for video calls you can use here as well: https://calendar.app.google/8qt4a5Yw8NzQN9s17 Thank you again and I hope we can use your feedback to improve the onboarding and install process @chilly-nail-50429 šŸ˜„
e
This is wonderful detailed feedback Nicola, thanks so much!
c
1&2: I setup a machine dedicated to be a sandbox, using ubuntu server minimal. 3: • my first visit was: https://docs.openhands.dev/openhands/usage/agent-canvas/overview • which led me to: https://docs.openhands.dev/openhands/usage/agent-canvas/backend-setup/docker My plan is to have a sandbox just running openhands (it's an old computer), which prevent my main computer to be impacted. I run host an LLM on that main computer with Ollama.
I still have a few weird behaviors (errors popping up in the top right) and when switching sessions and LLMs, the backend seems to crash sometimes. But once it ran, I was able to do sessions of up to 30 minutes without hiccups (haven't tried more than 30 minutes yet). I tried using other AI agents, and for the time being, I really like the results I get without having the need to get too involved. Pretty neat!
e
Thanks for the feedback, if you have any examples of the errors let us know! I'd be curious what you find the most unique about agent canvas vs. other tools you may have used?