Fix CLI and Headless docs for after release

[mamoodi]

End-user friendly description of the problem this fixes or functionality that this introduces

• Include this change in the Release Notes. If checked, you must provide an end-user friendly description for your change below

Fix CLI and Headless docs for after release

---

Give a summary of what the PR does, explaining any non-trivial design decisions

First version 0.18 needs to be released. Just prepping this up so we don't forget.

---

Link of any specific issues this addresses
#5535

---

To run this PR locally, use the following command:

docker run -it --rm   -p 3000:3000   -v /var/run/docker.sock:/var/run/docker.sock   --add-host host.docker.internal:host-gateway   -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:d41fd98-nikolaik   --name openhands-app-d41fd98   docker.all-hands.dev/all-hands-ai/openhands:d41fd98

[enyst]

This is different from CLI Mode, which is interactive, and better for active development.

Maybe we'd like it to be better, in time, but as of now I would remove that. Currently it seems that compared to main.py, cli.py:

• doesn't restore sessions
• doesn't allow naming sessions
• or most other cli parameters.

These are limitations that are on my TODO list to fix. FWIW, I am using main.py most of the time, and I definitely think it's "better". It's a subjective thing after all.

[mamoodi]

This is different from CLI Mode, which is interactive, and better for active development.

Maybe we'd like it to be better, in time, but as of now I would remove that. Currently it seems that compared to main.py, cli.py:

• doesn't restore sessions
• doesn't allow naming sessions
• or most other cli parameters.

These are limitations that are on my TODO list to fix. FWIW, I am using main.py most of the time, and I definitely think it's "better". It's a subjective thing after all.

I'll have to come back to this because I found more issues. Maybe you know more.

By passing the new .openhands-state through the -v param as I've done in the PR, both the Headless and CLI create a session that seem to always persist?
For example in headless if you give it a task to "create a hello world program in bash" it will do it.
If you then delete the file and re-run the same command, because previously it had already accomplished that task, it will not do it again because it says it was already done.

Robert tells me this isn't desired behavior. I'm guessing we need to pass something into the command so it knows it's a new task? Kind of confused how all of this works with all the new settings changes.

[enyst]

Thank you, mamoodi, that's a good point. Hmm. OK, let's give it some thought.

What we can always do, with main.py, is send it the command line parameter -n, which is a session name. So when I send something like main.py -t "some task" -n "bedrock1", it runs the task on a "bedrock1_4u7462...etc" event stream.

Then if I want a new session, I can do main.py -t "some task" -n "bedrock2", which runs the task on a different session name (thus different event stream, "bedrock2_blah-blah-a-hash). If I never used that before, it's a new, empty stream, there are no events to load.

[mamoodi]

That's interesting. I can test it out and see if that works. Thanks!

If it does work, it might be worth it to rejig the documentation for CLI and Headless a bit.
Basically add the changes I have in my current PR, but then at the end of CLI and Headless, add some documentation for some of the different flags you can pass to the command, such as the -n which ensures a unique session. Thoughts?

[enyst]

Yes, that's my thought as well.

[mamoodi]

@enyst with the changes done on main, the current changes in the PR work as expected. I'd like to get this in along with the release. Then will circle back and rejig the documentation. Thanks so much for your help here. I'm starting to understand it.