Update app.py

app.py

@@ -1,4 +1,4 @@
-# app.py (最终稳定版 - 使用 gr.Blocks)
+# app.py (最终确认版 - 使用 gr.Blocks)
 import gradio as gr
 from langchain.prompts import PromptTemplate
 from langchain_community.embeddings import HuggingFaceBgeEmbeddings
@@ -12,18 +12,21 @@ import time
 # --- 1. 配置 (保持不变) ---
 VECTOR_STORE_PATH = "vector_store"
 EMBEDDING_MODEL = "BAAI/bge-large-zh-v1.5"
+# 切换到 CapybaraHermes-2.5-Mistral-7B 模型
 GGUF_MODEL_REPO = "TheBloke/CapybaraHermes-2.5-Mistral-7B-GGUF"
+# 我们选择一个大小适中的4位量化版本
 GGUF_MODEL_FILE = "capybarahermes-2.5-mistral-7b.Q4_K_M.gguf"
 
 # --- 2. 加载RAG管道 (保持不变) ---
 def load_rag_chain():
-    # ... (这部分代码和之前完全一样，无需修改) ...
     print("开始加载RAG管道...")
     embeddings = HuggingFaceBgeEmbeddings(model_name=EMBEDDING_MODEL, model_kwargs={'device': 'cpu'}, encode_kwargs={'normalize_embeddings': True})
     if not os.path.exists(VECTOR_STORE_PATH): raise FileNotFoundError(f"错误：向量数据库 '{VECTOR_STORE_PATH}' 不存在！")
     vector_store = FAISS.load_local(VECTOR_STORE_PATH, embeddings, allow_dangerous_deserialization=True)
     model_path = hf_hub_download(repo_id=GGUF_MODEL_REPO, filename=GGUF_MODEL_FILE, local_dir="models")
     llm = LlamaCpp(model_path=model_path, n_gpu_layers=0, n_batch=512, n_ctx=4096, f16_kv=True, verbose=False)
+    
+    # 使用为Mistral模型优化的Prompt模板
     prompt_template = """<|im_start|>system
 You are a helpful assistant named "粤小智". Answer the user's question in Chinese based on the provided "Context".
 If the context is not sufficient, just say: "抱歉，关于您的问题，我的知识库暂时没有相关信息。". Do not make up answers.
@@ -41,10 +44,9 @@ Question:
     print("✅ RAG管道加载完毕！")
     return qa_chain
 
-# --- 3. Gradio应用逻辑 (修改以适配gr.Blocks) ---
+# --- 3. Gradio应用逻辑 (适配gr.Blocks) ---
 RAG_CHAIN = load_rag_chain()
 
-# history是Gradio自动管理的，格式为[ [user_msg1, bot_msg1], [user_msg2, bot_msg2], ... ]
 def user(user_message, history):
     # 将用户消息添加到聊天记录中，并返回一个空的输入框
     return "", history + [[user_message, None]]
@@ -58,7 +60,7 @@ def bot(history):
     result = RAG_CHAIN.invoke({"query": user_message})
     bot_message = result.get('result', "处理出错").strip()
     
-    # 我们模拟打字效果，让体验更好
+    # 模拟打字效果
     history[-1][1] = ""
     for character in bot_message:
         history[-1][1] += character
@@ -90,6 +92,6 @@ with gr.Blocks(theme=gr.themes.Soft(), css="footer {display: none !important}")
         bot, chatbot, chatbot
     )
 
-# 使用最简单的启动方式，但加入queue()来处理打字效果
+# 使用queue()来处理流式（打字效果）输出
 demo.queue()
 demo.launch()