local-llm-server/llm_server/llm/info.py

import requests

from llm_server import opts


def get_running_model():
    # TODO: cache the results for 1 min so we don't have to keep calling the backend
    # TODO: only use one try/catch

    if opts.mode == 'oobabooga':
        try:
            backend_response = requests.get(f'{opts.backend_url}/api/v1/model', timeout=3, verify=opts.verify_ssl)
            r_json = backend_response.json()
            return r_json['result'], None
        except Exception as e:
            return False, e
    elif opts.mode == 'hf-textgen':
        try:
            backend_response = requests.get(f'{opts.backend_url}/info', verify=opts.verify_ssl)
            r_json = backend_response.json()
            return r_json['model_id'].replace('/', '_'), None
        except Exception as e:
            return False, e
    elif opts.mode == 'vllm':
        try:
            backend_response = requests.get(f'{opts.backend_url}/model', timeout=3, verify=opts.verify_ssl)
            r_json = backend_response.json()
            return r_json['model'], None
        except Exception as e:
            return False, e
    else:
        raise Exception