1. 非流式输出模式

需要安装依赖：

pip install gradio

pip install langchain

pip install langchain-openai

核心代码：

import gradio as gr
from langchain.chat_models import init_chat_model
from langchain_core.messages import SystemMessage, HumanMessage, AIMessage

# 核心配置
API_KEY = "XXXXXXX"        # 阿里百炼 API_KEY
BASE_URL = "XXXXXXXXXX"    # 对应模型的 URL
MODEL_NAME = "qwen3.8-max"
TEMPERATURE = 0.7

# 使用init_chat_model初始化模型，兼容OpenAI兼容接口
llm = init_chat_model(
    model=MODEL_NAME,
    model_provider="openai",
    api_key=API_KEY,
    base_url=BASE_URL,
    temperature=TEMPERATURE
)

def chat_with_ai(message, history):
    """
    history：Gradio messages格式：[{"role":"user","content":"xxx"}, {"role":"assistant","content":"xxx"}]
    转成langchain消息对象
    """
    msg_list = [SystemMessage(content="你是一个 helpful 的助手。")]

    # 遍历messages格式历史，转为LangChain对象
    for msg in history:
        role = msg["role"]
        content = msg["content"]
        if role == "user":
            msg_list.append(HumanMessage(content=content))
        elif role == "assistant":
            msg_list.append(AIMessage(content=content))

    msg_list.append(HumanMessage(content=message))

    try:
        resp = llm.invoke(msg_list)
        return resp.content
    except Exception as e:
        return f"调用失败：{str(e)}"

with gr.Blocks(title="AI聊天助手") as demo:
    gr.Markdown("# AI聊天助手")

    chatbot = gr.Chatbot(height=500)
    msg_box = gr.Textbox(placeholder="输入你的问题...", label="输入")
    clear_btn = gr.Button("清空对话")

    def on_submit(user_text, chat_history):
        if not user_text.strip():
            return "", chat_history

        reply = chat_with_ai(user_text, chat_history)
        # 追加 messages字典格式，不再用元组！
        chat_history.append({"role": "user", "content": user_text})
        chat_history.append({"role": "assistant", "content": reply})
        return "", chat_history

    msg_box.submit(
        fn=on_submit,
        inputs=[msg_box, chatbot],
        outputs=[msg_box, chatbot]
    )

    clear_btn.click(fn=lambda: [], inputs=None, outputs=chatbot, queue=False)

if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", share=False)

效果图：

2. 流式输出模式

import gradio as gr
from langchain.chat_models import init_chat_model
from langchain_core.messages import SystemMessage, HumanMessage, AIMessage

# 核心配置
API_KEY = "xxxx"
BASE_URL = "xxxx"
MODEL_NAME = "qwen3.8-max"
TEMPERATURE = 0.7

# 初始化模型
llm = init_chat_model(
    model=MODEL_NAME,
    model_provider="openai",
    api_key=API_KEY,
    base_url=BASE_URL,
    temperature=TEMPERATURE
)


def chat_stream_generator(message, history):
    """流式生成器，返回逐块文本"""
    msg_list = [SystemMessage(content="你是一个 helpful 的助手。")]

    for msg in history:
        role = msg["role"]
        content = msg["content"]
        if role == "user":
            msg_list.append(HumanMessage(content=content))
        elif role == "assistant":
            msg_list.append(AIMessage(content=content))

    msg_list.append(HumanMessage(content=message))

    full_content = ""
    try:
        for chunk in llm.stream(msg_list):
            full_content += chunk.content
            yield full_content
    except Exception as e:
        yield f"调用失败：{str(e)}"


with gr.Blocks(title="AI聊天助手") as demo:
    gr.Markdown("# AI聊天助手")

    chatbot = gr.Chatbot(height=500)
    msg_box = gr.Textbox(placeholder="输入你的问题...", label="输入")
    clear_btn = gr.Button("清空对话")

    def on_submit(user_text, chat_history):
        if not user_text.strip():
            return "", chat_history

        # 先把用户消息加入历史
        chat_history.append({"role": "user", "content": user_text})
        # 初始化空的AI回复
        chat_history.append({"role": "assistant", "content": ""})

        # 流式迭代
        for partial_text in chat_stream_generator(user_text, chat_history[:-2]):
            # 更新最后一条assistant消息
            chat_history[-1]["content"] = partial_text
            yield "", chat_history

    msg_box.submit(
        fn=on_submit,
        inputs=[msg_box, chatbot],
        outputs=[msg_box, chatbot]
    )

    clear_btn.click(fn=lambda: [], inputs=None, outputs=chatbot, queue=False)


if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", share=False)