LangChain使用入门

安装LangChain

uv add langchain
uv add langchain-openai

本地启动一个大模型

参考文章juejin.cn/post/762965…

(base) gillbert@pop-os:~$ conda activate llm
(llm) gillbert@pop-os:~$ cd models/
(llm) gillbert@pop-os:~/models$ vllm serve DeepSeek-R1-Distill-Qwen-1.5B --tensor-parallel-size 1 --gpu-memory-utilization 0.7 --max-model-len 4096 --host 0.0.0.0 --port 8000 --api-key 123456

连接模型

from langchain.chat_models import init_chat_model

model = init_chat_model(
    model="DeepSeek-R1-Distill-Qwen-1.5B",
    model_provider="openai",
    api_key="123456",
    base_url="http://192.168.0.23:8000/v1"
)

print(model.invoke("你是谁").content)

print("*" * 50)

输出如下

流式输出

from langchain.chat_models import init_chat_model
import logging

from langchain_core.language_models import BaseChatModel

logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)

def init_llm_client() -> BaseChatModel:
    model = init_chat_model(
        model="DeepSeek-R1-Distill-Qwen-1.5B",
        model_provider="openai",
        api_key="123456",
        base_url="http://192.168.0.23:8000/v1"
    )

    return model

def main():
    try:
        chat = init_llm_client()
        logger.info("LLM客户端初始化成功")
        # 流式输出
        response_stream = chat.stream("介绍一下日语的动词变形")
        for chunk in response_stream:
            print(chunk.content, end="")
    except Exception as e:
        logger.error(e)


if __name__ == "__main__":
    main()