francedot
commented
3 days ago This feature is on our roadmap, but unfortunately, I can’t provide a specific timeline. It will require some modifications to the current Docker setup. We welcome contributions, so if you feel up to the task, here's some guidance to help you get started:
- In the current architecture, both the client and server processes are hosted in the same Docker container. Specifically, the VM runs within the Docker container and executes the server (VM Controller). You can refer to this diagram for the components involved. To achieve this, you will need to decouple the client and server, and run the server directly on your Windows host.
- To run the benchmark tasks, you'll first need to set up your environment with the necessary tools and programs. The setup.ps1 PowerShell script provides what’s required to do so. You might want to generalize the script so it works both on your host and within the current Docker setup.
- Next, you will need to run the Python server (VM Controller) component directly on your Windows host. To do this, create a new Conda environment on your Windows system and run
pip install -r requirementsfrom this requirements file. - The client process will need to point to the running py server on your localhost. You can find the references that need to be updated in the client here.
- The user folder, where programs are installed and recordings are stored, will need to be abstracted to support both Docker and the host setup. The relevant references to update are here.
Calvvnono
Hello,
I would like to directly deploy the
NaviAgent on my current operating system (without the need for virtual machines or cloud setups) and test a specific custom task. My goal is to observe the agent's performance in real-time on my local environment.Could you please clarify if this is feasible, and if so, what steps should I follow to achieve this?
Any guidance on this process would be greatly appreciated.
Thank you!