Code-WebUI / app.py
Povlakek-ai's picture
Update app.py
7b7e8c9 verified
Raw
History Blame Contribute Delete
1.82 kB
import os
import gradio as gr
from huggingface_hub import InferenceClient
# 1. 从环境变量中读取 Token
hf_token = os.environ.get("HF_TOKEN")
# 2. 定义模型 ID
MODEL_ID = "meta-llama/Llama-3.2-3B-Instruct"
# 3. 初始化 Hugging Face 推理客户端
client = InferenceClient(model=MODEL_ID, token=hf_token)
# 4. 修复后的聊天逻辑
def chat_fn(message, history):
messages = []
# 兼容标准的 Gradio 历史记录格式:[[user_msg1, ai_msg1], [user_msg2, ai_msg2], ...]
for user_msg, assistant_msg in history:
if user_msg:
messages.append({"role": "user", "content": user_msg})
if assistant_msg:
messages.append({"role": "assistant", "content": assistant_msg})
# 加上当前用户发送的新消息
messages.append({"role": "user", "content": message})
response = ""
try:
# 调用流式 API
for message_chunk in client.chat_completion(
messages=messages,
max_tokens=1024,
stream=True,
temperature=0.7,
top_p=0.9
):
token = message_chunk.choices[0].delta.content
if token:
response += token
yield response
except Exception as e:
yield f"⚠️ 出错了!可能是 API 暂时繁忙。错误信息: {str(e)}"
# 5. 构建 Gradio 聊天界面(删除了引发错误的 type="messages")
demo = gr.ChatInterface(
fn=chat_fn,
title=f"🤖 我的专属大模型聊天室 ({MODEL_ID.split('/')[-1]})",
description="基于 Hugging Face 免费 Serverless API 驱动,不占用本地硬件资源,速度极快!",
examples=["你好,请自我介绍一下。", "用 Python 写一个快速排序算法。"]
)
if __name__ == "__main__":
demo.launch()