An example of how it would work is that the agent would run in one container and it would generate code to be run in another container, or they could be run in the same container. This way the target container wouldn't need the stuff to interface to the AI to be there. Also the container with the interface to the AI would need to be bootstrapped, it seems - though this bootstrapping could be one-shotted.
An example of how it would work is that the agent would run in one container and it would generate code to be run in another container, or they could be run in the same container. This way the target container wouldn't need the stuff to interface to the AI to be there. Also the container with the interface to the AI would need to be bootstrapped, it seems - though this bootstrapping could be one-shotted.