hwhkit

Main function

• Connection
  • mqtt
• llm

Connection

yaml config

key_pairs:
  default_topic:
    algorithm: rsa
    public: '-----BEGIN PUBLIC KEY-----'
    private: '-----BEGIN PRIVATE KEY-----'
  default_topic_2:
    algorithm: aes
    private: '---------xxx-------'

Sync MQTT

import time
import signal
from hwhkit.connection.mqtt.client import MQTTClientManager

def main():
    default_topic = "default_topic"
    client_id = "test_mqtt_client"
    manager = MQTTClientManager(mqtt_config="mqtt_keys.yaml")
    manager.create_client(client_id=client_id, broker="broker.emqx.io", port=1883)
    manager.start_all_clients()

    @manager.subscribe(topic=default_topic)
    def handle_message(client, message: str):
        print(f"Received message from {client}: {message}")
        manager.publish(client_id, default_topic, f"Response from {client}")

    time.sleep(4)
    manager.publish(client_id=client_id, topic=default_topic, message="Hello from Client2")
    signal.pause()

if __name__ == '__main__':
    main()

Async MQTT

import asyncio
from hwhkit.connection.mqtt.async_client import MQTTAsyncClientManager, MQTTConfig
from hwhkit.utils import logger

async def main():
    configs = [
        MQTTConfig(client_id="client1", broker="broker.emqx.io", port=1883, username="user", password="pass"),
    ]
    default_topic = "default_topic"
    async with MQTTAsyncClientManager(mqtt_config="mqtt_keys.yaml") as manager:
        for config in configs:
            await manager.add_client(config)

        @manager.topic_handler(default_topic)
        async def topic_key(client, topic, message):
            logger.info(f"Received message on {topic} from {client}: {message}")
            await manager.publish("client1", default_topic, f"Response from {client}")

        await manager.run()

if __name__ == "__main__":
    asyncio.run(main())

LLM

Three steps to use models

Step1, llm_config.yaml

matter that needs attention

1. custom_model_name used for models.get_model_instance()
2. custom_model_name.name should specify the name of the model supported by the current company

models:
  openai:
    custom_model_name:
      name: "gpt-4o"
      short_name: "OIG4"
      company: "openai"
      max_input_token: 8100
      max_output_token: 2048
      top_p: 0.5
      top_k: 1
      temperature: 0.5
      input_token_fee_pm: 30.0
      output_token_fee_pm: 60.0
      train_token_fee_pm: 0.0
      keys:
        - name: "openai_key1"
        - name: "openai_key2"

  siliconflow:
    qw-72b-p:
      name: "Qwen/QVQ-72B-Preview"
      short_name: "QW-72B-P"
      company: "siliconflow"
      max_input_token: 8100
      max_output_token: 2048
      top_p: 0.5
      top_k: 1
      temperature: 0.5
      input_token_fee_pm: 30.0
      output_token_fee_pm: 60.0
      train_token_fee_pm: 0.0
      keys:
        - name: "siliconflow_1"

Step2, llm_keys.yaml

1. The keys name of the model in llm_config.yaml corresponds to llm_keys.yaml one by one

keys:
  openai_key1: "xx"
  openai_key2: "xx"
  anthropic_key1: "your_anthropic_api_key_1"
  anthropic_key2: "your_anthropic_api_key_2"

Step3, load models

from hwhkit.llm.config import load_models_from_yaml
from hwhkit.llm import LLMClient

async def main():
  
    # The first method
    models = load_models_from_yaml(config_file="examples/llm_config.yaml", keys_file="examples/llm_keys.yaml")
    print(models.list_models())

    resp = await models.get_model_instance("gpt-4o").chat("who r u?")
    print(resp)

    # The second method
    client = LLMClient(config_file="llm_config.yaml", keys_file="llm_keys.yaml")
    print(client.list_models())
    resp = await client.chat("qw-72b-p", "who r u?", system_prompt="")
    print(resp)
    async for chunk in client.chat_stream("qw-72b-p", "who r u?", system_prompt=""):
        print(chunk)

if __name__ == '__main__':
    import asyncio
    asyncio.run(main())