Backend flag support

[ilopezluna]

Add a --backend flag to docker model run and docker model list.
For the run case we use the backend provided as prefix.
For the list case we use the OpenAI's endpoint in case of --backend openai

In both cases you must specify OPENAI_API_KEY

[xenoscopic]

LGTM, certainly the logic and design looks great, just had a few minor nit suggestions.

[doringeman]

Suggested change

	c.Flags().StringVar(&backend, "backend", "", "Specify the backend to use (llama.cpp, openai)")
	c.Flags().StringVar(&backend, "backend", "", "Specify the backend to use ("+
	strings.Join(slices.Collect(maps.Keys(ValidBackends)), ", ")+")")

Even add a function to return the keys as it'd be used in at least 3 places - validateBackend, flag for list, flag for run.

func ValidBackendsKeys() []string {
	return slices.Collect(maps.Keys(ValidBackends))
}

[ilopezluna]

addressed here

[doringeman]

LGTM!

[ericcurtin]

Haven't installed this yet, but what's the difference between openai and llama.cpp in this context? Since llama-server from llama.cpp provides an openai-compatible endpoint?

[ilopezluna]

Haven't installed this yet, but what's the difference between openai and llama.cpp in this context? Since llama-server from llama.cpp provides an openai-compatible endpoint?

The backend defines the engine responsible for running the inference. When the backend is set to llama.cpp, the inference is executed using llama.cpp.
A new openai backend its been added, allowing inference to be performed via OpenAI.
This new flag is meant to define the engine used to do inference:

We can include other backends as long as they support OpenAI compatible API.