Create app.py

app.py

@@ -0,0 +1,58 @@
+import os
+import gradio as gr
+from langchain_huggingface import HuggingFaceEndpoint, HuggingFaceEmbeddings
+from langchain_community.vectorstores import FAISS
+from langchain_community.document_loaders import PyPDFLoader, TextLoader, Docx2txtLoader
+from langchain_text_splitters import RecursiveCharacterTextSplitter
+from langchain.chains import RetrievalQA
+
+# 1. 接入顶级大脑 API
+llm = HuggingFaceEndpoint(
+    repo_id="Qwen/Qwen2.5-7B-Instruct",
+    huggingfacehub_api_token=os.getenv("HF_TOKEN")
+)
+
+# 2. 向量化模型（负责把文字变成AI能懂的坐标）
+embeddings = HuggingFaceEmbeddings(model_name="BAAI/bge-small-zh-v1.5")
+
+def process_files(files):
+    all_docs = []
+    for file in files:
+        # 根据文件后缀选择加载器
+        if file.name.endswith('.pdf'):
+            loader = PyPDFLoader(file.name)
+        elif file.name.endswith('.docx'):
+            loader = Docx2txtLoader(file.name)
+        else:
+            loader = TextLoader(file.name)
+        all_docs.extend(loader.load())
+    
+    # 3. 智能切片（防止文档太长 AI 记不住）
+    text_splitter = RecursiveCharacterTextSplitter(chunk_size=600, chunk_overlap=100)
+    split_docs = text_splitter.split_documents(all_docs)
+    
+    # 4. 存入临时向量库
+    vectorstore = FAISS.from_documents(split_docs, embeddings)
+    return vectorstore
+
+# 5. 问答函数
+def predict(message, history, file_output):
+    if not file_output:
+        return "请先上传 PDF/Word/TXT 文档，我才能开启私有大脑模式。"
+    
+    # 处理文件并建立检索链
+    vectorstore = process_files(file_output)
+    qa_chain = RetrievalQA.from_chain_type(llm=llm, retriever=vectorstore.as_retriever())
+    
+    response = qa_chain.invoke(message)
+    return response["result"]
+
+# 6. 界面设计
+demo = gr.ChatInterface(
+    predict,
+    additional_inputs=[gr.File(file_count="multiple", label="上传私有文档 (.pdf, .docx, .txt)")],
+    title="我的全能私有大脑",
+    description="上传文件后，我可以基于文档内容回答任何专业问题。"
+)
+
+demo.launch()