[Bug]: "The action has not been executed"

[avi12]

Is there an existing issue for the same bug?

• I have checked the existing issues.

Describe the bug and reproduction steps

I noticed that it started occurring after the agent attempted to execute

grep -rn 'types/types' /workspace/extension

OpenHands Installation

Docker command in README

OpenHands Version

0.18

Operating System

Windows

Logs, Errors, Screenshots, and Additional Context

Some of the log that was copied from the Docker output: scratch.log

[enyst]

At the point you show in the screenshot, are you able to enter a message in the chat, to continue?

The message "The action has not been executed" only says that the previous action encountered a runtime error, a timeout higher than 120 seconds, so it hasn't done what it was supposed to do.

The log shows that the client waits for docker and docker is unable to give a response, until it hits the timeout after 120 seconds.

I suggest to run docker under WSL. Have you tried that, how does it work on WSL?

[avi12]

At the point you show in the screenshot, are you able to enter a message in the chat, to continue?

I stopped the conversation and did other experimentations so I have no idea
Generally, I'd expect that if the agent encounters an error, it needs to reuse the last prompt to try to recover from that error

I suggest to run docker under WSL. Have you tried that, how does it work on WSL?

There is no need to run anything via WSL as Docker is platform-agnostic

[enyst]

At the point you show in the screenshot, are you able to enter a message in the chat, to continue?

I stopped the conversation and did other experimentations so I have no idea Generally, I'd expect that if the agent encounters an error, it needs to reuse the last prompt to try to recover from that error

There two types of errors, errors from which the agent can recover and try to fix/continue, and errors that are bad enough that the task has failed. We then stop execution, to not spin forever, and return control to the user.

A RuntimeError in the runtime (sandbox / sandbox communication) is one of those.

I suggest to run docker under WSL. Have you tried that, how does it work on WSL?

There is no need to run anything via WSL as Docker is platform-agnostic

That's the marketing version! 😅 In reality, running on Windows may have issues that don't exist on other platforms. Sorry, I cannot reproduce some problem running grep, it works.

I'm not going to be able to help on Windows, but FWIW I see an error like:

2025-01-03T07:31:45.788705108Z requests.exceptions.ReadTimeout: HTTPConnectionPool(host='host.docker.internal', port=39067): Read timed out. (read timeout=125)

If this is the issue, you may want to increase the timeout in the sandbox. In config.toml, you can add:

...
[sandbox]
timeout=240

The default is 120 seconds.

Or, if you're running with docker run, as an env var, -e SANDBOX_TIMEOUT=240.