server

var (
	Host string
	Port string

	Vocab *ml.Vocab
	Model *llama.Model

	Ctx    *llama.Context
	Params *llama.ModelParams

	MaxPods     int64
	RunningPods int64 // number of pods running right now in parallel

	// TODO: Background watcher which will make waiting jobs obsolete after some deadline
	Jobs  map[string]*Job     // all seen jobs in any state
	Queue map[string]struct{} // queue of job IDs waiting for start
)