Deploy

Installation

First install the docker and the nvidia-container-toolkit.

Then you can install the Pody server from pip:

pip install "pody[server]"

TIP

If you run the server as a non-root user, please add the user to the docker group.
For details, refer to the Docker documentation (linux-postinstall).

Start the server

The server stores the data in $PODY_HOME directory, which is by default ~/.pody.

To start the server:

pody-server --port 8799

Then by setting appropriate environment variables, you can use the pody-cli to interact with the server.

TIP

You can use systemd to start the server on boot. For details, refer to the start on boot guide.

Managements

Users and quotas

To manage users:

pody-user ...

To manage user quotas:

pody-quota ...

Please refer to --help for more information.

Configurations

The server configuration is stored in $PODY_HOME/config.toml file. There are comments in the file to help you understand the options.

For example, to manage images, you should first pull or build the image, then specify the images to expose to the client by editing the [[images]] section in the $PODY_HOME/config.toml file.

For a more detailed introduction, see the configuration guide.

Optional: Using docker network

If you want to allow containers to communicate with each other in a dedicated network (with DNS resolution), you can create a user-defined docker network (e.g. pody-net) and specify it in the container configuration.

docker network create pody-net

then, set the network option in the congfiguration file:

network = "pody-net"