```

feat(config): 添加RAG/Embedding配置选项

- 新增EMBEDDING_MODEL配置项，默认值为"all-MiniLM-L6-v2"
- 用于支持RAG服务的嵌入模型配置

feat(database): 增强MySQL数据库初始化功能

- 实现数据库自动创建功能，若数据库不存在则自动创建
- 使用临时连接在不指定数据库的情况下执行CREATE DATABASE语句
- 支持utf8mb4字符集和排序规则设置

refactor(excel): 优化Excel表创建逻辑

- 将表创建方式从ORM模型改为原生SQL语句
- 提高异步操作的兼容性
- 增加自动时间戳字段(created_at, updated_at)

feat(rag): 增强RAG服务嵌入模型错误处理

- 添加嵌入模型加载异常处理机制
- 当配置的模型加载失败时自动回退到默认模型
- 改进日志记录，提供更详细的初始化信息
```

backend/app/config.py

@@ -29,6 +29,9 @@ class Settings(BaseSettings):
     LLM_BASE_URL: str = "https://api.minimax.chat"
     LLM_MODEL_NAME: str = "MiniMax-Text-01"
 
+    # ==================== RAG/Embedding 配置 ====================
+    EMBEDDING_MODEL: str = "all-MiniLM-L6-v2"
+
     # ==================== Supabase 配置 ====================
     SUPABASE_URL: str = ""
     SUPABASE_ANON_KEY: str = ""

backend/app/core/database/mysql.py

@@ -72,6 +72,26 @@ class MySQLDB:
     async def init_db(self):
         """初始化数据库，创建所有表"""
         try:
+            # 先创建数据库（如果不存在）
+            from sqlalchemy import text
+            db_name = settings.MYSQL_DATABASE
+            # 连接时不指定数据库来创建数据库
+            temp_url = (
+                f"mysql+aiomysql://{settings.MYSQL_USER}:{settings.MYSQL_PASSWORD}"
+                f"@{settings.MYSQL_HOST}:{settings.MYSQL_PORT}/"
+                f"?charset={settings.MYSQL_CHARSET}"
+            )
+            from sqlalchemy.ext.asyncio import create_async_engine
+            temp_engine = create_async_engine(temp_url, echo=False)
+            try:
+                async with temp_engine.connect() as conn:
+                    await conn.execute(text(f"CREATE DATABASE IF NOT EXISTS `{db_name}` CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci"))
+                    await conn.commit()
+                logger.info(f"MySQL 数据库 {db_name} 创建或已存在")
+            finally:
+                await temp_engine.dispose()
+
+            # 然后创建表
             async with self.async_engine.begin() as conn:
                 await conn.run_sync(Base.metadata.create_all)
             logger.info("MySQL 数据库表初始化完成")

backend/app/services/excel_storage_service.py

@@ -203,13 +203,19 @@ class ExcelStorageService:
                     "type": col_type
                 })
 
-            # 创建表
-            model_class = self._create_table_model(table_name, df.columns, column_types)
-
-            # 创建表结构
+            # 创建表 - 使用原始 SQL 以兼容异步
             logger.info(f"正在创建MySQL表: {table_name}")
-            async with self.mysql_db.get_session() as session:
-                model_class.__table__.create(session.bind, checkfirst=True)
+            from sqlalchemy import text
+            sql_columns = ["id INT AUTO_INCREMENT PRIMARY KEY"]
+            for col in df.columns:
+                col_name = self._sanitize_column_name(col)
+                col_type = column_types.get(col, "TEXT")
+                sql_type = "INT" if col_type == "INTEGER" else "FLOAT" if col_type == "FLOAT" else "DATETIME" if col_type == "DATETIME" else "TEXT"
+                sql_columns.append(f"`{col_name}` {sql_type}")
+            sql_columns.append("created_at DATETIME DEFAULT CURRENT_TIMESTAMP")
+            sql_columns.append("updated_at DATETIME DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP")
+            create_sql = text(f"CREATE TABLE IF NOT EXISTS `{table_name}` ({', '.join(sql_columns)})")
+            await self.mysql_db.execute_raw_sql(str(create_sql))
             logger.info(f"MySQL表创建完成: {table_name}")
 
             # 插入数据

backend/app/services/rag_service.py

@@ -40,9 +40,17 @@ class RAGService:
     def _init_embeddings(self):
         """初始化嵌入模型"""
         if self.embedding_model is None:
-            self.embedding_model = SentenceTransformer(settings.EMBEDDING_MODEL)
+            model_name = getattr(settings, 'EMBEDDING_MODEL', 'all-MiniLM-L6-v2')
+            try:
+                self.embedding_model = SentenceTransformer(model_name)
                 self._dimension = self.embedding_model.get_sentence_embedding_dimension()
-            logger.info(f"RAG 嵌入模型初始化完成: {settings.EMBEDDING_MODEL}, 维度: {self._dimension}")
+                logger.info(f"RAG 嵌入模型初始化完成: {model_name}, 维度: {self._dimension}")
+            except Exception as e:
+                logger.warning(f"嵌入模型 {model_name} 加载失败，使用默认模型: {e}")
+                # 使用轻量级默认模型
+                self.embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
+                self._dimension = self.embedding_model.get_sentence_embedding_dimension()
+                logger.info(f"RAG 嵌入模型使用默认: all-MiniLM-L6-v2, 维度: {self._dimension}")
 
     def _init_vector_store(self):
         """初始化向量存储"""