pax_global_header00006660000000000000000000000064150666450100014515gustar00rootroot0000000000000052 comment=489b5ba4113a610a93091db52b58f67fd6663b1c langchain_project/000077500000000000000000000000001506664501000145335ustar00rootroot00000000000000langchain_project/.idea/000077500000000000000000000000001506664501000155135ustar00rootroot00000000000000langchain_project/.idea/langchain_project.iml000066400000000000000000000004341506664501000216710ustar00rootroot00000000000000 langchain_project/README.md000066400000000000000000000006731506664501000160200ustar00rootroot00000000000000# middle25projects reports collected for Middleware class project @2025 ## Report Naming Convension ``` 2023201001  ┃  ┃  ┗ experiment report (2023201001.md / 2023201001.tex / ...) ┃ ┗ src ┃ ┗ scripts(experiment scripts in Matlab / Python / ...) ┗ dataset(if over 100MB, use netdisk url) ┗ paper(OPTIONAL, share paper with info.ruc@163.com on overleaf) ``` [How to submit](./Submission-Howto.md)langchain_project/Submission-Howto.md000066400000000000000000000013241506664501000203060ustar00rootroot00000000000000# Submission Steps - create account on openi & login into openi.pcl.ac.cn - fork a copy of info-ruc/middle25projects - download and install git tool on your local computer - git config --global user.name "$yourname$" & git config --global user.email "$yourname@yourserver.com$" - git clone https://openi.pcl.ac.cn/$yourname$/middle25projects.git - make folders structures following Report Naming Convention in README.md - enter folders and edit reports (student_no.md / student_no.tex) and project files (datasets & scripts) - git add --all - git commit -m "description of edit" - git push (enter user name & password if needed) - create a pull request on openi to merge your commit with info-ruc/middle25projectslangchain_project/super_simple_langchain_web.py000066400000000000000000000063731506664501000224660ustar00rootroot00000000000000""" 超级简单的LangChain网页演示 这个脚本展示了LangChain的最基本功能，使用本地模型，并通过Gradio提供Web界面。 """ from langchain.prompts import PromptTemplate from langchain_huggingface import HuggingFacePipeline from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline import torch import gradio as gr # 检测可用设备 device = "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu" print(f"使用设备: {device}") # 加载支持中文的文本生成模型 model_name = "uer/gpt2-chinese-cluecorpussmall" # 一个支持中文的GPT-2模型 print(f"加载模型: {model_name}...") tokenizer = AutoTokenizer.from_pretrained(model_name) model = AutoModelForCausalLM.from_pretrained(model_name).to(device) # 创建HuggingFace pipeline pipe = pipeline( "text-generation", # 使用text-generation任务 model=model, tokenizer=tokenizer, max_new_tokens=50, device=device ) # 创建LangChain LLM llm = HuggingFacePipeline(pipeline=pipe) # 创建提示模板 prompt = PromptTemplate( input_variables=["topic"], template="请简要描述一下{topic}是什么：" ) # 使用LangChain的管道操作符 chain = prompt | llm # 定义Gradio处理函数 def generate_description(topic): if not topic: return "请输入一个主题！" # 格式化提示 formatted_prompt = prompt.format(topic=topic) print(f"处理提示: {formatted_prompt}") # 使用LangChain链调用模型 try: result = chain.invoke({"topic": topic}) # 清理结果，去除原始提示 cleaned_result = result.replace(formatted_prompt, "").strip() return f"关于 '{topic}' 的描述:\n\n{cleaned_result}" except Exception as e: return f"错误: {str(e)}" # 创建Gradio界面 with gr.Blocks(title="LangChain 中文演示") as demo: gr.Markdown("# 🤖 LangChain 中文演示") gr.Markdown("输入一个主题，获取AI生成的简短描述。") with gr.Row(): with gr.Column(): topic_input = gr.Textbox( label="主题", placeholder="输入一个主题（例如：人工智能、气候变化等）", lines=2 ) generate_button = gr.Button("生成描述", variant="primary") with gr.Column(): output = gr.Textbox( label="生成的描述", lines=8, interactive=False ) # 添加中文示例 gr.Examples( examples=["人工智能", "气候变化", "太空探索", "量子计算", "医疗健康"], inputs=topic_input ) # 设置事件处理 generate_button.click( fn=generate_description, inputs=topic_input, outputs=output ) topic_input.submit( fn=generate_description, inputs=topic_input, outputs=output ) gr.Markdown("### 工作原理") gr.Markdown(""" 本演示使用了： - 🔗 **LangChain** 用于提示模板和链式调用 - 🤗 **Hugging Face** 提供多语言模型 (DistilBERT) - 🖥️ **Gradio** 提供网页界面 """) # 启动Gradio应用 if __name__ == "__main__": demo.launch()