when running terminal commands let the user run them and return the feedback to the agent, it very much often uses the wrong version of python, sometimes I prefer to run it using a docker compose, so I have to cut it consuming a fast credit request and stopping the answer half way
Hey, have you tried adding this to your AI rules? You can also specify the required Python version and the command the agent should execute.
too much if a hassle would be just better if could edit the command before hitting the run command button directly in the generated chat window
Then disable the auto-run mode to edit the command.
it is off what I want is, instead of consuming too many fast requests just to get the right command why can’t just edit it on the agent chat tab, it’s more optimized this way
for example in the screenshot it is using just regular python it would’ve been much better if I could’ve just edited that command to use a python from a venv instead. because at this point the command won’t work and it will keep retrying and installing packages, …, or I cancel it and loose a fast request while the job is half done. it could be fixed by specifying it in the prompt or project rules but doesn’t fully fix it, the best approach is just to manually edit the command on the fly.
I meant to completely disable auto-run mode so that you wouldn’t need to cancel your chat, since the command is waiting for your confirmation
when the command is waiting for confirmation if you click cancel it will cancel the entire chat. the feature I am requesting is the ability to edit the command before hitting the ‘run command’ button
Don’t cancel it. Instead, edit the command and then click run command.
I don’t seem to be able to edit the command! how can I do that?
It might not work for you, but it works for me. You can just try editing the command in the opened terminal, just like I did in the video.
just updated to the latest version it works now, thanks
one cool feature would also be to skip the command in case I don’t it to pursue with that specific though process, sometimes I find it wanting to commit some code before running some tests … I know there is a skip option but I believe you only get after it fails the command a couple times it would be great if we get ‘cancel’, ‘run’, and ‘skip’ buttons from the first try