Override Model Name in llama.cpp HTTP Server /models API

[RodriMora]

Hi!

Llama.cpp HTTP Server seems to output the full path of the model

I have tested to override the model name with --override-kv general.name=str:gpt-3.5-turbo but it doesn't work as the api will output the full path of the file and not the general.name.

One solution might be to change this behavior and make llama.cpp server api output the id as the kv general.name instead of the path of the model

Another solution is to add a parameter like --override-model-name to change this

For example, currently this:
curl http://192.168.1.151:5000/v1/models
Will output:
{"object":"list","data":[{"id":"/home/ubuntuai/models/WizardLM-2-8x22B-Q4_K_M.gguf","object":"model","created":1721242854,"owned_by":"llamacpp","meta":{"vocab_type":1,"n_vocab":32000,"n_ctx_train":65536,"n_embd":6144,"n_params":140620634112,"size":85591511040}}]}%

I think it's better by default to output the general.name or give the possibility to override the name passing a parameter.

[ELigoP]

Looks like '--alias ...' parameter works for server binary.

https://github.com/ggerganov/llama.cpp/blob/master/examples/server/README.md