def main()

in arctic_inference/dynasor/vllm_server.py [0:0]

• 41 lines of code
• 8 McCabe index (conditional complexity)

def main():
    
    register_signal_handlers()

    parser = FlexibleArgumentParser(
        description="Arctic Inference Dynasor Proxy w/ vLLM Server"
    )
    parser = make_arg_parser(parser)
    args = parser.parse_args()

    vllm_port = args.port
    openai_port = vllm_port
    args.port = vllm_port + 1
    vllm_port = args.port
    validate_parsed_serve_args(args)

    # Start the server in a separate process
    server_process = multiprocessing.Process(target=run_vllm_server, args=(args,))

    
    openai_config = openai_server.parse_args([
        "--port", str(openai_port),
        "--target-base-url", f"http://localhost:{vllm_port}",
    ])
    print(f"Starting OpenAI server: {openai_config}")
    proxy_server = multiprocessing.Process(
        target=run_openai_server, 
        args=(openai_config,)
    )
    



    server_process.start()

    # Wait for the server to be healthy
    if check_health(vllm_port):
        print(f"Server is healthy and running on port {args.port}")
    else:
        print("Server failed to become healthy within the timeout period")
        server_process.terminate()
        server_process.join()
        sys.exit(1)

    proxy_server.start()

    # Wait for the OpenAI server to be healthy
    if check_health(openai_port):
        print(f"Dynasor Proxy server is healthy and running on port {openai_port}")
    else:
        print("Dynasor Proxy server failed to become healthy within the timeout period")
        proxy_server.terminate()
        proxy_server.join()
        sys.exit(1)

    # Keep the main process running and handle signals
    while server_process.is_alive() or proxy_server.is_alive(): 
        server_process.join(timeout=3.0)
        proxy_server.join(timeout=3.0)