-
-
Notifications
You must be signed in to change notification settings - Fork 1
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[QUESTION]: How to resume a saved session? #144
Comments
I mean, how to resume any of the saved sessions, not just the last one. |
Could you apply commit 1d4f692 and check using the conversation ID like http://localhost:3000/conversations/ae299e4431a54b52b82a45602ec81a13? |
Is it possible to keep the sandboxes created and resume them when the session is resumed? |
Yes, 2200b21 |
I'm testing the latest version of your repo, and I'm unable to reconnect to a previous conversation. The VS Code web endpoint also is not working. |
Solved. |
Remaining issues: VS Code web is not working (in all sessions): |
It's unstable. To reproduce:
Result: The sandbox container is started, but OH cannot connect to it. This are the processes running inside the sandbox:
|
I confirmed that each container uses a different HTTP for the "PID 8" ( But the front-end says: |
After stopping the sandbox (to reproduce the bug), sometimes the sandbox it is autorestarted and generates this logs:
BTW, even when OH lost connection to the sandbox (web terminal doesn't work), the Web UI (VS Code) endpoint works fine: Thus, the problem is only the "Open in VS Code" button in the frontend. |
Added persist sandbox for each conversation. |
[EDIT] Nevermind. I rebuilt everything and it seems to work now. I also added this to
|
I'm still unable to run your last repo version. BTW, why are the LLM UI settings lost after restarting the OH Container? This is my config:
Logs:
|
If so, Timeout error will occur. Is |
In Github Codespaces, does |
|
Yes, It's based on Ubuntu 20.04.6 |
So, this is not needed. right? |
|
How long does it take? Could you increase the timeout?
or could you remove this 7ee6e4a |
I added the repo (based on the dev Dockerfile), because Anyway, I believe the
But the problem is that VS Code Docker extension seems not to support the HEREDOC tags used in the dev Dockerfile. Remember I'm not using Codespaces, but trying to use DevContainer as a dev environment on Windows. On the other hand, trying to build the dev contianer on windows (using the composer file) also fails. How did you build your dev container on Windows? |
I didn't build it yet. |
Does it happen due to the line endings? moby/buildkit#3511 |
It's not time timeout:
|
That helps:
Now I have to find out why the endpoint answers with a |
Did you curl without |
[EDIT] Yes. I thought I got out of storage space on my laptop, because it was working earlier. |
|
After this, a new conversation works? |
Upstream is not using Selenium. I am manually installing it for quick workaround here. It shouldn't be in the |
Are you releasing PRs to upstream? |
Not planned yet as it is using playwright and shouldn't playwright impl be improved? |
I'm not sure, but upstream was working fine the other day for web browsing tasks. I mean if you are releasing PRs in general (other features, not selenium), so that upstream developers don't reinvent the wheel and help to maintain/enhance your implementations. I saw they were working on the front end for listing conversations. |
Waiting for All-Hands-AI#2782 to be reopened. |
Maybe better isolate the "Auto Mode" related commits and create a PR with all other more important commits to concentrate the development on a single code base. It's common to have some personal commits on a separate personal branch until upstream feels the need to merge or solves the issue in a different or more general way. IMO, the support of mulitple conversations and the ability to resume them is gold and they need it urgently upstream. |
20 PRs were cleaned up. |
Feel free to create a PR using these changes. |
I would like to, but there are so many changes and I'm still not enough familiarized with everything. I would probably miss some dependency. I would suggest you to 1) create a PR for resuming multiple conversations before they diverge, and after that, 2) take to look at the upstream roadmap and the wanted PRs section in Slack to see what else may fit. I can help you with the PRs revision. Considering the high number of collaborators, I'm pretty sure the upstream managers are overwhelmed and focusing on the project roadmap. |
Is this doing the same? Upstream already implanted a conversations panel and the option to destroy conversations. |
Yes. With this change and All-Hands-AI#6080 when |
|
My friend,: let's better slowly get back to upstream. Otherwise we will be writing and testing code that will be overwritten by thousand of contributors on the upstream repo. The good thing is that you have been merging from upstream. |
FYI, upstream automode: |
To release a PR with the persistent_sessions changes to upstream, we should:
Other features/PRs should be handled in similar way in separate branches with separate PRs. If you want I can help you with that. Otherwise I feel we are spending a lot of time writing and testing code that will be overwritten soon, plus your effort in merging redundant code every day. What do you think? |
I tested upstream and it's almost able to resume sessions after a reboot. The number of new PRs every day is crazy. |
It's only for CLI. |
It's working now. I was struggling with All-Hands-AI#6148. |
I miss the "beep" (when OH completes a task) in upstream. |
It's disabled? [EDIT] Ah, All-Hands-AI@80ba1fd |
To enable the conversations panel, they used localStorage: Another easter egg :-) |
All-Hands-AI#5726 Issue
@kripper It will be resumed automatically.
The text was updated successfully, but these errors were encountered: