Update README.md

no timespan on gpu status

Slim (#3)

add /ping endpoint and remove gpu locks if stale formore than 1 min

add controller to reserve and release gpus

remove 'islocked' since it's redundant with modelInUse

kestral it up why not

adjust update on the channel

no idea if it'll work but added openai controller

change from storing just a bool to a small object

gpu manager uses dataservice for redis operations

go back to simpler response handling

chunked/not chunked working now

healtcheck isn't special.still trying to get streams workign

looping comes back chunked when its not