fix: 修复缓存权限和根路径404问题

- 在Dockerfile中创建并设置HuggingFace缓存目录权限
- 设置HF_HOME等环境变量指向可写目录
- 在AI模型中明确指定缓存目录
- 修复根路径404错误
- 解决Permission denied: '/.cache'问题

Dockerfile

@@ -1,5 +1,5 @@
-# Dockerfile - 更新版本
-FROM python:3.11-slim 
+# Dockerfile - 修复缓存权限
+FROM python:3.11-slim
 
 # 设置工作目录
 WORKDIR /app
@@ -8,6 +8,9 @@ WORKDIR /app
 ENV PYTHONPATH=/app
 ENV PYTHONUNBUFFERED=1
 ENV PIP_NO_CACHE_DIR=1
+ENV HF_HOME=/app/.cache/huggingface  # 设置HF缓存目录
+ENV TRANSFORMERS_CACHE=/app/.cache/huggingface
+ENV HF_DATASETS_CACHE=/app/.cache/huggingface
 
 # 安装系统依赖
 RUN apt-get update && apt-get install -y \
@@ -16,6 +19,11 @@ RUN apt-get update && apt-get install -y \
     build-essential \
     && rm -rf /var/lib/apt/lists/*
 
+# 创建缓存目录并设置权限
+RUN mkdir -p /app/.cache/huggingface && \
+    chmod 755 /app/.cache && \
+    chmod 755 /app/.cache/huggingface
+
 # 升级 pip 到最新版本
 RUN pip install --upgrade pip
 
@@ -26,9 +34,10 @@ RUN pip install --no-cache-dir -r requirements.txt
 # 复制项目文件
 COPY . .
 
-#hf_cli authenticate
+# 在构建时进行HF认证（如果token可用）
 ARG Assitant_tocken
 RUN if [ -n "$Assitant_tocken" ]; then \
+        export HF_HOME=/app/.cache/huggingface && \
         python -c "from huggingface_hub import login; login('$Assitant_tocken', add_to_git_credential=False)"; \
     fi
 

app.py

@@ -1,4 +1,3 @@
-# app.py - 正确的版本
 from fastapi import FastAPI, HTTPException
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import JSONResponse
@@ -27,7 +26,7 @@ app.add_middleware(
 # --- 全局服务实例 ---
 try:
     log.info("🔄 开始初始化 Travel Assistant 服务...")
-    assistant = TravelAssistant()  # 这里才是正确的模型加载方式
+    assistant = TravelAssistant()
     SERVICE_READY = True
     log.info("🚀 FastAPI 应用启动成功，服务已就绪。")
 except Exception as e:
@@ -36,4 +35,125 @@ except Exception as e:
     log.critical(f"💥 FATAL: 服务初始化失败: {e}")
     log.critical(f"💥 错误详情: {traceback.format_exc()}")
 
-# ... 其余的API端点代码
+# --- API 端点 ---
+@app.get("/")
+async def root():
+    """根路径 - 提供基本信息"""
+    ai_available = False
+    if SERVICE_READY and assistant and hasattr(assistant, 'ai_model'):
+        ai_available = assistant.ai_model.is_available()
+    
+    return {
+        "message": "Travel Assistant API",
+        "version": "2.0.0",
+        "status": "ready" if SERVICE_READY else "degraded", 
+        "ai_model_status": "available" if ai_available else "unavailable",
+        "docs": "/docs",
+        "health": "/health"
+    }
+
+@app.get("/health")
+async def health_check():
+    """健康检查端点"""
+    if SERVICE_READY and assistant:
+        ai_available = assistant.ai_model.is_available() if hasattr(assistant.ai_model, 'is_available') else False
+        return {
+            "status": "healthy" if ai_available else "degraded",
+            "service_ready": True,
+            "ai_model_available": ai_available,
+            "components": {
+                "config": True,
+                "knowledge_base": len(assistant.kb.knowledge) > 0 if hasattr(assistant, 'kb') else False,
+                "ai_model": ai_available,
+                "session_manager": True
+            }
+        }
+    else:
+        return JSONResponse(
+            status_code=503,
+            content={
+                "status": "unhealthy", 
+                "service_ready": False,
+                "ai_model_available": False
+            }
+        )
+
+@app.post("/api/chat", response_model=ChatResponse)
+async def chat_endpoint(request: ChatRequest):
+    """主要的聊天API端点"""
+    if not SERVICE_READY or not assistant:
+        raise HTTPException(
+            status_code=503,
+            detail="Service Unavailable: Backend assistant failed to initialize."
+        )
+    
+    try:
+        log.info(f"收到聊天请求: {request.message[:50]}...")
+        
+        reply, session_id, status_info, history = assistant.chat(
+            request.message, 
+            request.session_id,
+            request.history or []
+        )
+        
+        log.info(f"聊天响应生成成功，会话ID: {session_id}")
+        
+        return ChatResponse(
+            reply=reply,
+            session_id=session_id,
+            status_info=status_info,
+            history=history
+        )
+        
+    except Exception as e:
+        log.error(f"❌ Chat endpoint error: {e}", exc_info=True)
+        raise HTTPException(
+            status_code=500,
+            detail="Internal Server Error: Failed to process chat request."
+        )
+
+@app.post("/api/reset")
+async def reset_session(session_id: str):
+    """重置会话端点"""
+    if not SERVICE_READY or not assistant:
+        raise HTTPException(
+            status_code=503,
+            detail="Service Unavailable"
+        )
+    
+    try:
+        assistant.session_manager.reset(session_id)
+        log.info(f"会话重置成功: {session_id}")
+        return {
+            "message": "Session reset successfully", 
+            "session_id": session_id,
+            "status": "success"
+        }
+    except Exception as e:
+        log.error(f"❌ Reset session error: {e}")
+        raise HTTPException(
+            status_code=500,
+            detail="Failed to reset session"
+        )
+
+# --- 全局异常处理 ---
+@app.exception_handler(Exception)
+async def global_exception_handler(request, exc):
+    """全局异常处理器"""
+    log.error(f"❌ 未处理的异常: {exc}", exc_info=True)
+    return JSONResponse(
+        status_code=500,
+        content={"detail": "Internal server error", "message": "请稍后重试"}
+    )
+
+# HuggingFace Spaces 和本地运行配置
+if __name__ == "__main__":
+    import uvicorn
+    log.info("🔧 本地开发模式启动...")
+    uvicorn.run(
+        app, 
+        host="0.0.0.0", 
+        port=7860,
+        reload=True,
+        log_level="info"
+    )

modules/ai_model.py

@@ -17,66 +17,60 @@ class AIModel:
         self.processor = None
         self._initialize_model()
 
-    # modules/ai_model.py - 添加调试版本
-    def _authenticate_hf(self):
-        try:
-            # 检查所有可能的环境变量
-            assitant_token = os.getenv("Assitant_tocken")
-            hf_token = os.getenv("HUGGINGFACE_HUB_TOKEN") 
-            hf_token_alt = os.getenv("HF_TOKEN")
-        
-            log.info("=== 环境变量调试 ===")
-            log.info(f"Assitant_tocken: {'存在' if assitant_token else '不存在'}")
-            log.info(f"HUGGINGFACE_HUB_TOKEN: {'存在' if hf_token else '不存在'}")
-            log.info(f"HF_TOKEN: {'存在' if hf_token_alt else '不存在'}")
+    def _setup_cache_dirs(self):
+        """设置缓存目录"""
+        cache_dir = "/app/.cache/huggingface"
+        os.makedirs(cache_dir, exist_ok=True)
         
-            # 使用找到的token
-            token_to_use = assitant_token or hf_token or hf_token_alt
+        # 设置环境变量
+        os.environ["HF_HOME"] = cache_dir
+        os.environ["TRANSFORMERS_CACHE"] = cache_dir
+        os.environ["HF_DATASETS_CACHE"] = cache_dir
         
-            if token_to_use:
-                log.info(f"使用token: {token_to_use[:10]}...")
-                login(token=token_to_use, add_to_git_credential=False)
+        log.info(f"设置缓存目录: {cache_dir}")
+
+    def _authenticate_hf(self):
+        try:
+            # 从环境变量获取HF token
+            hf_token = os.getenv("Assitant_tocken")
+            
+            if hf_token:
+                cache_dir = "/app/.cache/huggingface"
+                login(token=hf_token, add_to_git_credential=False)
                 log.info("✅ HuggingFace 认证成功")
-                return token_to_use
             else:
-                log.error("❌ 未找到任何有效的 HuggingFace token")
-                return None
-            
+                log.warning("⚠️ 未找到 HuggingFace token，可能无法访问受限模型")
+                
         except Exception as e:
             log.error(f"❌ HuggingFace 认证失败: {e}")
-            return None
+
 
     def _initialize_model(self):
+        """初始化Gemma模型 - 基于官方调用方式"""
         try:
             log.info(f"正在加载模型: {self.model_name}")
-        
-            # 先进行认证并获取token
-            token = self._authenticate_hf()
-        
-            if not token:
-                log.error("❌ 无法获取有效token，模型加载失败")
-                self.model = None
-                self.processor = None
-                return
-        
+
+            hf_token = self._authenticate_hf()
+
+            cache_dir = "/app/.cache/huggingface"
+            
             self.model = Gemma3nForConditionalGeneration.from_pretrained(
                 self.model_name,
                 device_map="auto",
                 torch_dtype=torch.bfloat16,
-                trust_remote_code=True,
-                token=token,
-                use_auth_token=token  # 双重保险
+                cache_dir=cache_dir,
+                trust_remote_code=True
             ).eval()
-        
+            
             self.processor = AutoProcessor.from_pretrained(
                 self.model_name,
                 trust_remote_code=True,
-                token=token,
-                use_auth_token=token
+                cache_dir=cache_dir,
+                tocken=hf_token
             )
-        
+            
             log.info("✅ Gemma AI 模型初始化成功")
-        
+            
         except Exception as e:
             log.error(f"❌ Gemma AI 模型初始化失败: {e}", exc_info=True)
             self.model = None