Feat socket io

[tofarr]

Using SocketIO as the Underlying Communication Protocol

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

---

Migration / Refactor to use SocketIO as the communication protocol instead of raw Websockets

• Better Error Tolerance
• Better Clusterability (Using Redis)

Testing Scenarios:

Starting a new session should open a single websocket

Refreshing the page should allow the session to continue uninterrupted (I refreshed the page then asked a follow on question):

Interrupting the socket without refreshing the page should allow the session to continue uninterrupted (You can see in the screenshot that I used the developer console to close the socket, and a new one instantly opened to replace it. I asked a follow up question with no issue):

Many of the Use cases for this PR (Particularly related to Redis are mostly applicable in the SAAS environment, where Pods may cease to exist mid session and we want to minimize user disruption. The logic when dealing with multiple pods is more complex than with a single pod!

---

Fixes #5151

---

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:56ca2ed-nikolaik   --name openhands-app-56ca2ed   docker.all-hands.dev/all-hands-ai/openhands:56ca2ed

[diwu-sf]

Will this make #5019 easier to implement by decoupling the Session from the websocket?

[rbren]

should we os.Exit here or anything?

[tofarr]

No - if there are any other graceful shutdown tasks that might interfere with those - ending the loop should be sufficient.

[rbren]

Only bit that'd be nice to tackle is that between SIGTERM and SIGKILL the agent appears as STOPPED, even though it's still working during the grace period. We can tackle that in a follow-on though, this is working well as-is