123

- Keep user's RAG service integration for faster extraction
- Add remote's word_ai_service support
- Preserve user's parallel extraction and field header optimizations

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

backend/.env.example

@@ -29,9 +29,14 @@ REDIS_URL="redis://localhost:6379/0"
 
 # ==================== LLM AI 配置 ====================
 # 大语言模型 API 配置
-LLM_API_KEY="your_api_key_here"
-LLM_BASE_URL=""
-LLM_MODEL_NAME=""
+# 支持 OpenAI 兼容格式 (DeepSeek, 智谱 GLM, 阿里等)
+# 智谱 AI (Zhipu AI) GLM 系列:
+#   - 模型: glm-4-flash (快速文本模型), glm-4 (标准), glm-4-plus (高性能)
+#   - API: https://open.bigmodel.cn
+#   - API Key: https://open.bigmodel.cn/usercenter/apikeys
+LLM_API_KEY="ca79ad9f96524cd5afc3e43ca97f347d.cpiLLx2oyitGvTeU"
+LLM_BASE_URL="https://open.bigmodel.cn/api/paas/v4"
+LLM_MODEL_NAME="glm-4v-plus"
 
 # ==================== Supabase 配置 ====================
 # Supabase 项目配置

backend/=3.0.0

@@ -0,0 +1,38 @@
+Requirement already satisfied: sentence-transformers in c:\python312\lib\site-packages (2.2.2)
+Requirement already satisfied: transformers<5.0.0,>=4.6.0 in c:\python312\lib\site-packages (from sentence-transformers) (4.57.6)
+Requirement already satisfied: tqdm in c:\python312\lib\site-packages (from sentence-transformers) (4.66.1)
+Requirement already satisfied: torch>=1.6.0 in c:\python312\lib\site-packages (from sentence-transformers) (2.10.0)
+Requirement already satisfied: torchvision in c:\python312\lib\site-packages (from sentence-transformers) (0.25.0)
+Requirement already satisfied: numpy in c:\python312\lib\site-packages (from sentence-transformers) (1.26.2)
+Requirement already satisfied: scikit-learn in c:\python312\lib\site-packages (from sentence-transformers) (1.8.0)
+Requirement already satisfied: scipy in c:\python312\lib\site-packages (from sentence-transformers) (1.16.3)
+Requirement already satisfied: nltk in c:\python312\lib\site-packages (from sentence-transformers) (3.9.3)
+Requirement already satisfied: sentencepiece in c:\python312\lib\site-packages (from sentence-transformers) (0.2.1)
+Requirement already satisfied: huggingface-hub>=0.4.0 in c:\python312\lib\site-packages (from sentence-transformers) (0.36.2)
+Requirement already satisfied: filelock in c:\python312\lib\site-packages (from huggingface-hub>=0.4.0->sentence-transformers) (3.25.2)
+Requirement already satisfied: fsspec>=2023.5.0 in c:\python312\lib\site-packages (from huggingface-hub>=0.4.0->sentence-transformers) (2026.2.0)
+Requirement already satisfied: packaging>=20.9 in c:\python312\lib\site-packages (from huggingface-hub>=0.4.0->sentence-transformers) (23.2)
+Requirement already satisfied: pyyaml>=5.1 in c:\python312\lib\site-packages (from huggingface-hub>=0.4.0->sentence-transformers) (6.0.1)
+Requirement already satisfied: requests in c:\python312\lib\site-packages (from huggingface-hub>=0.4.0->sentence-transformers) (2.31.0)
+Requirement already satisfied: typing-extensions>=3.7.4.3 in c:\python312\lib\site-packages (from huggingface-hub>=0.4.0->sentence-transformers) (4.15.0)
+Requirement already satisfied: sympy>=1.13.3 in c:\python312\lib\site-packages (from torch>=1.6.0->sentence-transformers) (1.14.0)
+Requirement already satisfied: networkx>=2.5.1 in c:\python312\lib\site-packages (from torch>=1.6.0->sentence-transformers) (3.6.1)
+Requirement already satisfied: jinja2 in c:\python312\lib\site-packages (from torch>=1.6.0->sentence-transformers) (3.1.6)
+Requirement already satisfied: setuptools in c:\python312\lib\site-packages (from torch>=1.6.0->sentence-transformers) (82.0.1)
+Requirement already satisfied: colorama in c:\python312\lib\site-packages (from tqdm->sentence-transformers) (0.4.6)
+Requirement already satisfied: regex!=2019.12.17 in c:\python312\lib\site-packages (from transformers<5.0.0,>=4.6.0->sentence-transformers) (2026.2.28)
+Requirement already satisfied: tokenizers<=0.23.0,>=0.22.0 in c:\python312\lib\site-packages (from transformers<5.0.0,>=4.6.0->sentence-transformers) (0.22.2)
+Requirement already satisfied: safetensors>=0.4.3 in c:\python312\lib\site-packages (from transformers<5.0.0,>=4.6.0->sentence-transformers) (0.7.0)
+Requirement already satisfied: click in c:\python312\lib\site-packages (from nltk->sentence-transformers) (8.3.1)
+Requirement already satisfied: joblib in c:\python312\lib\site-packages (from nltk->sentence-transformers) (1.5.3)
+Requirement already satisfied: threadpoolctl>=3.2.0 in c:\python312\lib\site-packages (from scikit-learn->sentence-transformers) (3.6.0)
+Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in c:\python312\lib\site-packages (from torchvision->sentence-transformers) (12.1.1)
+Requirement already satisfied: mpmath<1.4,>=1.1.0 in c:\python312\lib\site-packages (from sympy>=1.13.3->torch>=1.6.0->sentence-transformers) (1.3.0)
+Requirement already satisfied: MarkupSafe>=2.0 in c:\python312\lib\site-packages (from jinja2->torch>=1.6.0->sentence-transformers) (3.0.3)
+Requirement already satisfied: charset-normalizer<4,>=2 in c:\python312\lib\site-packages (from requests->huggingface-hub>=0.4.0->sentence-transformers) (3.4.6)
+Requirement already satisfied: idna<4,>=2.5 in c:\python312\lib\site-packages (from requests->huggingface-hub>=0.4.0->sentence-transformers) (3.11)
+Requirement already satisfied: urllib3<3,>=1.21.1 in c:\python312\lib\site-packages (from requests->huggingface-hub>=0.4.0->sentence-transformers) (2.6.3)
+Requirement already satisfied: certifi>=2017.4.17 in c:\python312\lib\site-packages (from requests->huggingface-hub>=0.4.0->sentence-transformers) (2026.2.25)
+
+[notice] A new release of pip is available: 24.2 -> 26.0.1
+[notice] To update, run: python.exe -m pip install --upgrade pip

backend/app/api/endpoints/ai_analyze.py

@@ -10,6 +10,8 @@ import os
 
 from app.services.excel_ai_service import excel_ai_service
 from app.services.markdown_ai_service import markdown_ai_service
+from app.services.template_fill_service import template_fill_service
+from app.services.word_ai_service import word_ai_service
 
 logger = logging.getLogger(__name__)
 
@@ -215,9 +217,12 @@ async def analyze_markdown(
             return result
 
         finally:
-            # 清理临时文件
-            if os.path.exists(tmp_path):
+            # 清理临时文件，确保在所有情况下都能清理
+            try:
+                if tmp_path and os.path.exists(tmp_path):
                     os.unlink(tmp_path)
+            except Exception as cleanup_error:
+                logger.warning(f"临时文件清理失败: {tmp_path}, error: {cleanup_error}")
 
     except HTTPException:
         raise
@@ -279,8 +284,12 @@ async def analyze_markdown_stream(
             )
 
         finally:
-            if os.path.exists(tmp_path):
+            # 清理临时文件，确保在所有情况下都能清理
+            try:
+                if tmp_path and os.path.exists(tmp_path):
                     os.unlink(tmp_path)
+            except Exception as cleanup_error:
+                logger.warning(f"临时文件清理失败: {tmp_path}, error: {cleanup_error}")
 
     except HTTPException:
         raise
@@ -289,7 +298,7 @@ async def analyze_markdown_stream(
         raise HTTPException(status_code=500, detail=f"流式分析失败: {str(e)}")
 
 
-@router.get("/analyze/md/outline")
+@router.post("/analyze/md/outline")
 async def get_markdown_outline(
     file: UploadFile = File(...)
 ):
@@ -323,9 +332,154 @@ async def get_markdown_outline(
             result = await markdown_ai_service.extract_outline(tmp_path)
             return result
         finally:
-            if os.path.exists(tmp_path):
+            # 清理临时文件，确保在所有情况下都能清理
+            try:
+                if tmp_path and os.path.exists(tmp_path):
                     os.unlink(tmp_path)
+            except Exception as cleanup_error:
+                logger.warning(f"临时文件清理失败: {tmp_path}, error: {cleanup_error}")
 
     except Exception as e:
         logger.error(f"获取 Markdown 大纲失败: {str(e)}")
         raise HTTPException(status_code=500, detail=f"获取大纲失败: {str(e)}")
+
+
+@router.post("/analyze/txt")
+async def analyze_txt(
+    file: UploadFile = File(...),
+):
+    """
+    上传并使用 AI 分析 TXT 文本文件，提取结构化数据
+
+    将非结构化文本转换为结构化表格数据，便于后续填表使用
+
+    Args:
+        file: 上传的 TXT 文件
+
+    Returns:
+        dict: 分析结果，包含结构化表格数据
+    """
+    if not file.filename:
+        raise HTTPException(status_code=400, detail="文件名为空")
+
+    file_ext = file.filename.split('.')[-1].lower()
+    if file_ext not in ['txt', 'text']:
+        raise HTTPException(
+            status_code=400,
+            detail=f"不支持的文件类型: {file_ext}，仅支持 .txt"
+        )
+
+    try:
+        # 读取文件内容
+        content = await file.read()
+
+        # 保存到临时文件
+        with tempfile.NamedTemporaryFile(mode='wb', suffix='.txt', delete=False) as tmp:
+            tmp.write(content)
+            tmp_path = tmp.name
+
+        try:
+            logger.info(f"开始 AI 分析 TXT 文件: {file.filename}")
+
+            # 使用 template_fill_service 的 AI 分析方法
+            result = await template_fill_service.analyze_txt_with_ai(
+                content=content.decode('utf-8', errors='replace'),
+                filename=file.filename
+            )
+
+            if result:
+                logger.info(f"TXT AI 分析成功: {file.filename}")
+                return {
+                    "success": True,
+                    "filename": file.filename,
+                    "structured_data": result
+                }
+            else:
+                logger.warning(f"TXT AI 分析返回空结果: {file.filename}")
+                return {
+                    "success": False,
+                    "filename": file.filename,
+                    "error": "AI 分析未能提取到结构化数据",
+                    "structured_data": None
+                }
+
+        finally:
+            # 清理临时文件
+            if os.path.exists(tmp_path):
+                os.unlink(tmp_path)
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"TXT AI 分析过程中出错: {str(e)}")
+        raise HTTPException(status_code=500, detail=f"分析失败: {str(e)}")
+
+
+# ==================== Word 文档 AI 解析 ====================
+
+@router.post("/analyze/word")
+async def analyze_word(
+    file: UploadFile = File(...),
+    user_hint: str = Query("", description="用户提示词，如'请提取表格数据'")
+):
+    """
+    使用 AI 解析 Word 文档，提取结构化数据
+
+    适用于从非结构化的 Word 文档中提取表格数据、键值对等信息
+
+    Args:
+        file: 上传的 Word 文件
+        user_hint: 用户提示词
+
+    Returns:
+        dict: 包含结构化数据的解析结果
+    """
+    if not file.filename:
+        raise HTTPException(status_code=400, detail="文件名为空")
+
+    file_ext = file.filename.split('.')[-1].lower()
+    if file_ext not in ['docx']:
+        raise HTTPException(
+            status_code=400,
+            detail=f"不支持的文件类型: {file_ext}，仅支持 .docx"
+        )
+
+    try:
+        # 保存上传的文件
+        content = await file.read()
+        suffix = f".{file_ext}"
+        with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as tmp:
+            tmp.write(content)
+            tmp_path = tmp.name
+
+        try:
+            # 使用 AI 解析 Word 文档
+            result = await word_ai_service.parse_word_with_ai(
+                file_path=tmp_path,
+                user_hint=user_hint or "请提取文档中的所有结构化数据，包括表格、键值对等"
+            )
+
+            if result.get("success"):
+                return {
+                    "success": True,
+                    "filename": file.filename,
+                    "result": result
+                }
+            else:
+                return {
+                    "success": False,
+                    "filename": file.filename,
+                    "error": result.get("error", "AI 解析失败"),
+                    "result": None
+                }
+
+        finally:
+            # 清理临时文件
+            if os.path.exists(tmp_path):
+                os.unlink(tmp_path)
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"Word AI 分析过程中出错: {str(e)}")
+        raise HTTPException(status_code=500, detail=f"分析失败: {str(e)}")

backend/app/api/endpoints/documents.py

@@ -23,6 +23,52 @@ logger = logging.getLogger(__name__)
 router = APIRouter(prefix="/upload", tags=["文档上传"])
 
 
+# ==================== 辅助函数 ====================
+
+async def update_task_status(
+    task_id: str,
+    status: str,
+    progress: int = 0,
+    message: str = "",
+    result: dict = None,
+    error: str = None
+):
+    """
+    更新任务状态，同时写入 Redis 和 MongoDB
+
+    Args:
+        task_id: 任务ID
+        status: 状态
+        progress: 进度
+        message: 消息
+        result: 结果
+        error: 错误信息
+    """
+    meta = {"progress": progress, "message": message}
+    if result:
+        meta["result"] = result
+    if error:
+        meta["error"] = error
+
+    # 尝试写入 Redis
+    try:
+        await redis_db.set_task_status(task_id, status, meta)
+    except Exception as e:
+        logger.warning(f"Redis 任务状态更新失败: {e}")
+
+    # 尝试写入 MongoDB（作为备用）
+    try:
+        await mongodb.update_task(
+            task_id=task_id,
+            status=status,
+            message=message,
+            result=result,
+            error=error
+        )
+    except Exception as e:
+        logger.warning(f"MongoDB 任务状态更新失败: {e}")
+
+
 # ==================== 请求/响应模型 ====================
 
 class UploadResponse(BaseModel):
@@ -77,6 +123,17 @@ async def upload_document(
     task_id = str(uuid.uuid4())
 
     try:
+        # 保存任务记录到 MongoDB（如果 Redis 不可用时仍能查询）
+        try:
+            await mongodb.insert_task(
+                task_id=task_id,
+                task_type="document_parse",
+                status="pending",
+                message=f"文档 {file.filename} 已提交处理"
+            )
+        except Exception as mongo_err:
+            logger.warning(f"MongoDB 保存任务记录失败: {mongo_err}")
+
         content = await file.read()
         saved_path = file_service.save_uploaded_file(
             content,
@@ -122,6 +179,17 @@ async def upload_documents(
     saved_paths = []
 
     try:
+        # 保存任务记录到 MongoDB
+        try:
+            await mongodb.insert_task(
+                task_id=task_id,
+                task_type="batch_parse",
+                status="pending",
+                message=f"已提交 {len(files)} 个文档处理"
+            )
+        except Exception as mongo_err:
+            logger.warning(f"MongoDB 保存批量任务记录失败: {mongo_err}")
+
         for file in files:
             if not file.filename:
                 continue
@@ -159,9 +227,9 @@ async def process_document(
     """处理单个文档"""
     try:
         # 状态: 解析中
-        await redis_db.set_task_status(
+        await update_task_status(
             task_id, status="processing",
-            meta={"progress": 10, "message": "正在解析文档"}
+            progress=10, message="正在解析文档"
         )
 
         # 解析文档
@@ -172,9 +240,9 @@ async def process_document(
             raise Exception(result.error or "解析失败")
 
         # 状态: 存储中
-        await redis_db.set_task_status(
+        await update_task_status(
             task_id, status="processing",
-            meta={"progress": 30, "message": "正在存储数据"}
+            progress=30, message="正在存储数据"
         )
 
         # 存储到 MongoDB
@@ -191,9 +259,9 @@ async def process_document(
 
         # 如果是 Excel，存储到 MySQL + AI生成描述 + RAG索引
         if doc_type in ["xlsx", "xls"]:
-            await redis_db.set_task_status(
+            await update_task_status(
                 task_id, status="processing",
-                meta={"progress": 50, "message": "正在存储到MySQL并生成字段描述"}
+                progress=50, message="正在存储到MySQL并生成字段描述"
             )
 
             try:
@@ -215,9 +283,9 @@ async def process_document(
 
         else:
             # 非结构化文档
-            await redis_db.set_task_status(
+            await update_task_status(
                 task_id, status="processing",
-                meta={"progress": 60, "message": "正在建立索引"}
+                progress=60, message="正在建立索引"
             )
 
             # 如果文档中有表格数据，提取并存储到 MySQL + RAG
@@ -238,36 +306,33 @@ async def process_document(
             await index_document_to_rag(doc_id, original_filename, result, doc_type)
 
         # 完成
-        await redis_db.set_task_status(
+        await update_task_status(
             task_id, status="success",
-            meta={
-                "progress": 100,
-                "message": "处理完成",
-                "doc_id": doc_id,
-                "result": {
+            progress=100, message="处理完成",
+            result={
                 "doc_id": doc_id,
                 "doc_type": doc_type,
                 "filename": original_filename
             }
-            }
         )
 
         logger.info(f"文档处理完成: {original_filename}, doc_id: {doc_id}")
 
     except Exception as e:
         logger.error(f"文档处理失败: {str(e)}")
-        await redis_db.set_task_status(
+        await update_task_status(
             task_id, status="failure",
-            meta={"error": str(e)}
+            progress=0, message="处理失败",
+            error=str(e)
         )
 
 
 async def process_documents_batch(task_id: str, files: List[dict]):
     """批量处理文档"""
     try:
-        await redis_db.set_task_status(
+        await update_task_status(
             task_id, status="processing",
-            meta={"progress": 0, "message": "开始批量处理"}
+            progress=0, message="开始批量处理"
         )
 
         results = []
@@ -318,37 +383,43 @@ async def process_documents_batch(task_id: str, files: List[dict]):
                 results.append({"filename": file_info["filename"], "success": False, "error": str(e)})
 
             progress = int((i + 1) / len(files) * 100)
-            await redis_db.set_task_status(
+            await update_task_status(
                 task_id, status="processing",
-                meta={"progress": progress, "message": f"已处理 {i+1}/{len(files)}"}
+                progress=progress, message=f"已处理 {i+1}/{len(files)}"
             )
 
-        await redis_db.set_task_status(
+        await update_task_status(
             task_id, status="success",
-            meta={"progress": 100, "message": "批量处理完成", "results": results}
+            progress=100, message="批量处理完成",
+            result={"results": results}
         )
 
     except Exception as e:
         logger.error(f"批量处理失败: {str(e)}")
-        await redis_db.set_task_status(
+        await update_task_status(
             task_id, status="failure",
-            meta={"error": str(e)}
+            progress=0, message="批量处理失败",
+            error=str(e)
         )
 
 
 async def index_document_to_rag(doc_id: str, filename: str, result: ParseResult, doc_type: str):
-    """将非结构化文档索引到 RAG"""
+    """将非结构化文档索引到 RAG（使用分块索引）"""
     try:
         content = result.data.get("content", "")
         if content:
+            # 将完整内容传递给 RAG 服务自动分块索引
             rag_service.index_document_content(
                 doc_id=doc_id,
-                content=content[:5000],
+                content=content,  # 传递完整内容，由 RAG 服务自动分块
                 metadata={
                     "filename": filename,
                     "doc_type": doc_type
-                }
+                },
+                chunk_size=500,  # 每块 500 字符
+                chunk_overlap=50  # 块之间 50 字符重叠
             )
+            logger.info(f"RAG 索引完成: {filename}, doc_id={doc_id}")
     except Exception as e:
         logger.warning(f"RAG 索引失败: {str(e)}")
 

backend/app/api/endpoints/health.py

@@ -19,26 +19,43 @@ async def health_check() -> Dict[str, Any]:
     返回各数据库连接状态和应用信息
     """
     # 检查各数据库连接状态
-    mysql_status = "connected"
-    mongodb_status = "connected"
-    redis_status = "connected"
+    mysql_status = "unknown"
+    mongodb_status = "unknown"
+    redis_status = "unknown"
 
     try:
         if mysql_db.async_engine is None:
             mysql_status = "disconnected"
-    except Exception:
+        else:
+            # 实际执行一次查询验证连接
+            from sqlalchemy import text
+            async with mysql_db.async_engine.connect() as conn:
+                await conn.execute(text("SELECT 1"))
+            mysql_status = "connected"
+    except Exception as e:
+        logger.warning(f"MySQL 健康检查失败: {e}")
         mysql_status = "error"
 
     try:
         if mongodb.client is None:
             mongodb_status = "disconnected"
-    except Exception:
+        else:
+            # 实际 ping 验证
+            await mongodb.client.admin.command('ping')
+            mongodb_status = "connected"
+    except Exception as e:
+        logger.warning(f"MongoDB 健康检查失败: {e}")
         mongodb_status = "error"
 
     try:
-        if not redis_db.is_connected:
+        if not redis_db.is_connected or redis_db.client is None:
             redis_status = "disconnected"
-    except Exception:
+        else:
+            # 实际执行 ping 验证
+            await redis_db.client.ping()
+            redis_status = "connected"
+    except Exception as e:
+        logger.warning(f"Redis 健康检查失败: {e}")
         redis_status = "error"
 
     return {

backend/app/api/endpoints/tasks.py

@@ -1,13 +1,13 @@
 """
 任务管理 API 接口
 
-提供异步任务状态查询
+提供异步任务状态查询和历史记录
 """
 from typing import Optional
 
 from fastapi import APIRouter, HTTPException
 
-from app.core.database import redis_db
+from app.core.database import redis_db, mongodb
 
 router = APIRouter(prefix="/tasks", tags=["任务管理"])
 
@@ -23,20 +23,10 @@ async def get_task_status(task_id: str):
     Returns:
         任务状态信息
     """
+    # 优先从 Redis 获取
     status = await redis_db.get_task_status(task_id)
 
-    if not status:
-        # Redis不可用时，假设任务已完成（文档已成功处理）
-        # 前端轮询时会得到这个响应
-        return {
-            "task_id": task_id,
-            "status": "success",
-            "progress": 100,
-            "message": "任务处理完成",
-            "result": None,
-            "error": None
-        }
-
+    if status:
         return {
             "task_id": task_id,
             "status": status.get("status", "unknown"),
@@ -45,3 +35,82 @@ async def get_task_status(task_id: str):
             "result": status.get("meta", {}).get("result"),
             "error": status.get("meta", {}).get("error")
         }
+
+    # Redis 不可用时，尝试从 MongoDB 获取
+    mongo_task = await mongodb.get_task(task_id)
+    if mongo_task:
+        return {
+            "task_id": mongo_task.get("task_id"),
+            "status": mongo_task.get("status", "unknown"),
+            "progress": 100 if mongo_task.get("status") == "success" else 0,
+            "message": mongo_task.get("message"),
+            "result": mongo_task.get("result"),
+            "error": mongo_task.get("error")
+        }
+
+    # 任务不存在或状态未知
+    return {
+        "task_id": task_id,
+        "status": "unknown",
+        "progress": 0,
+        "message": "无法获取任务状态（Redis和MongoDB均不可用）",
+        "result": None,
+        "error": None
+    }
+
+
+@router.get("/")
+async def list_tasks(limit: int = 50, skip: int = 0):
+    """
+    获取任务历史列表
+
+    Args:
+        limit: 返回数量限制
+        skip: 跳过数量
+
+    Returns:
+        任务列表
+    """
+    try:
+        tasks = await mongodb.list_tasks(limit=limit, skip=skip)
+        return {
+            "success": True,
+            "tasks": tasks,
+            "count": len(tasks)
+        }
+    except Exception as e:
+        # MongoDB 不可用时返回空列表
+        return {
+            "success": False,
+            "tasks": [],
+            "count": 0,
+            "error": str(e)
+        }
+
+
+@router.delete("/{task_id}")
+async def delete_task(task_id: str):
+    """
+    删除任务
+
+    Args:
+        task_id: 任务ID
+
+    Returns:
+        是否删除成功
+    """
+    try:
+        # 从 Redis 删除
+        if redis_db._connected and redis_db.client:
+            key = f"task:{task_id}"
+            await redis_db.client.delete(key)
+
+        # 从 MongoDB 删除
+        deleted = await mongodb.delete_task(task_id)
+
+        return {
+            "success": True,
+            "deleted": deleted
+        }
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=f"删除任务失败: {str(e)}")

backend/app/api/endpoints/templates.py

@@ -23,6 +23,44 @@ logger = logging.getLogger(__name__)
 router = APIRouter(prefix="/templates", tags=["表格模板"])
 
 
+# ==================== 辅助函数 ====================
+
+async def update_task_status(
+    task_id: str,
+    status: str,
+    progress: int = 0,
+    message: str = "",
+    result: dict = None,
+    error: str = None
+):
+    """
+    更新任务状态，同时写入 Redis 和 MongoDB
+    """
+    from app.core.database import redis_db
+
+    meta = {"progress": progress, "message": message}
+    if result:
+        meta["result"] = result
+    if error:
+        meta["error"] = error
+
+    try:
+        await redis_db.set_task_status(task_id, status, meta)
+    except Exception as e:
+        logger.warning(f"Redis 任务状态更新失败: {e}")
+
+    try:
+        await mongodb.update_task(
+            task_id=task_id,
+            status=status,
+            message=message,
+            result=result,
+            error=error
+        )
+    except Exception as e:
+        logger.warning(f"MongoDB 任务状态更新失败: {e}")
+
+
 # ==================== 请求/响应模型 ====================
 
 class TemplateFieldRequest(BaseModel):
@@ -41,6 +79,7 @@ class FillRequest(BaseModel):
     source_doc_ids: Optional[List[str]] = None  # MongoDB 文档 ID 列表
     source_file_paths: Optional[List[str]] = None  # 源文档文件路径列表
     user_hint: Optional[str] = None
+    task_id: Optional[str] = None  # 可选的任务ID，用于任务历史跟踪
 
 
 class ExportRequest(BaseModel):
@@ -186,13 +225,51 @@ async def upload_joint_template(
                     parser = ParserFactory.get_parser(sf_path)
                     parse_result = parser.parse(sf_path)
                     if parse_result.success and parse_result.data:
+                        # 获取原始内容
+                        content = parse_result.data.get("content", "")[:5000] if parse_result.data.get("content") else ""
+
+                        # 获取标题（可能在顶层或structured_data内）
+                        titles = parse_result.data.get("titles", [])
+                        if not titles and parse_result.data.get("structured_data"):
+                            titles = parse_result.data.get("structured_data", {}).get("titles", [])
+                        titles = titles[:10] if titles else []
+
+                        # 获取表格数量（可能在顶层或structured_data内）
+                        tables = parse_result.data.get("tables", [])
+                        if not tables and parse_result.data.get("structured_data"):
+                            tables = parse_result.data.get("structured_data", {}).get("tables", [])
+                        tables_count = len(tables) if tables else 0
+
+                        # 获取表格内容摘要（用于 AI 理解源文档结构）
+                        tables_summary = ""
+                        if tables:
+                            tables_summary = "\n【文档中的表格】:\n"
+                            for idx, table in enumerate(tables[:5]):  # 最多5个表格
+                                if isinstance(table, dict):
+                                    headers = table.get("headers", [])
+                                    rows = table.get("rows", [])
+                                    if headers:
+                                        tables_summary += f"表格{idx+1}表头: {', '.join(str(h) for h in headers)}\n"
+                                    if rows:
+                                        tables_summary += f"表格{idx+1}前3行: "
+                                        for row_idx, row in enumerate(rows[:3]):
+                                            if isinstance(row, list):
+                                                tables_summary += " | ".join(str(c) for c in row) + "; "
+                                            elif isinstance(row, dict):
+                                                tables_summary += " | ".join(str(row.get(h, "")) for h in headers if headers) + "; "
+                                        tables_summary += "\n"
+
                         source_contents.append({
                             "filename": sf.filename,
                             "doc_type": sf_ext,
-                            "content": parse_result.data.get("content", "")[:5000] if parse_result.data.get("content") else "",
-                            "titles": parse_result.data.get("titles", [])[:10] if parse_result.data.get("titles") else [],
-                            "tables_count": len(parse_result.data.get("tables", [])) if parse_result.data.get("tables") else 0
+                            "content": content,
+                            "titles": titles,
+                            "tables_count": tables_count,
+                            "tables_summary": tables_summary
                         })
+                        logger.info(f"[DEBUG] source_contents built: filename={sf.filename}, content_len={len(content)}, titles_count={len(titles)}, tables_count={tables_count}")
+                        if tables_summary:
+                            logger.info(f"[DEBUG] tables_summary preview: {tables_summary[:300]}")
                 except Exception as e:
                     logger.warning(f"解析源文档失败 {sf.filename}: {e}")
 
@@ -206,6 +283,17 @@ async def upload_joint_template(
         # 3. 异步处理源文档到MongoDB
         task_id = str(uuid.uuid4())
         if source_file_info:
+            # 保存任务记录到 MongoDB
+            try:
+                await mongodb.insert_task(
+                    task_id=task_id,
+                    task_type="source_process",
+                    status="pending",
+                    message=f"开始处理 {len(source_file_info)} 个源文档"
+                )
+            except Exception as mongo_err:
+                logger.warning(f"MongoDB 保存任务记录失败: {mongo_err}")
+
             background_tasks.add_task(
                 process_source_documents,
                 task_id=task_id,
@@ -244,12 +332,10 @@ async def upload_joint_template(
 
 async def process_source_documents(task_id: str, files: List[dict]):
     """异步处理源文档，存入MongoDB"""
-    from app.core.database import redis_db
-
     try:
-        await redis_db.set_task_status(
+        await update_task_status(
             task_id, status="processing",
-            meta={"progress": 0, "message": "开始处理源文档"}
+            progress=0, message="开始处理源文档"
         )
 
         doc_ids = []
@@ -278,22 +364,24 @@ async def process_source_documents(task_id: str, files: List[dict]):
                 logger.error(f"源文档处理异常: {file_info['filename']}, error: {str(e)}")
 
             progress = int((i + 1) / len(files) * 100)
-            await redis_db.set_task_status(
+            await update_task_status(
                 task_id, status="processing",
-                meta={"progress": progress, "message": f"已处理 {i+1}/{len(files)}"}
+                progress=progress, message=f"已处理 {i+1}/{len(files)}"
             )
 
-        await redis_db.set_task_status(
+        await update_task_status(
             task_id, status="success",
-            meta={"progress": 100, "message": "源文档处理完成", "doc_ids": doc_ids}
+            progress=100, message="源文档处理完成",
+            result={"doc_ids": doc_ids}
         )
         logger.info(f"所有源文档处理完成: {len(doc_ids)}个")
 
     except Exception as e:
         logger.error(f"源文档批量处理失败: {str(e)}")
-        await redis_db.set_task_status(
+        await update_task_status(
             task_id, status="failure",
-            meta={"error": str(e)}
+            progress=0, message="源文档处理失败",
+            error=str(e)
         )
 
 
@@ -352,7 +440,27 @@ async def fill_template(
     Returns:
         填写结果
     """
+    # 生成或使用传入的 task_id
+    task_id = request.task_id or str(uuid.uuid4())
+
     try:
+        # 创建任务记录到 MongoDB
+        try:
+            await mongodb.insert_task(
+                task_id=task_id,
+                task_type="template_fill",
+                status="processing",
+                message=f"开始填表任务: {len(request.template_fields)} 个字段"
+            )
+        except Exception as mongo_err:
+            logger.warning(f"MongoDB 创建任务记录失败: {mongo_err}")
+
+        # 更新进度 - 开始
+        await update_task_status(
+            task_id, "processing",
+            progress=0, message="开始处理..."
+        )
+
         # 转换字段
         fields = [
             TemplateField(
@@ -365,17 +473,51 @@ async def fill_template(
             for f in request.template_fields
         ]
 
+        # 从 template_id 提取文件类型
+        template_file_type = "xlsx"  # 默认类型
+        if request.template_id:
+            ext = request.template_id.split('.')[-1].lower()
+            if ext in ["xlsx", "xls"]:
+                template_file_type = "xlsx"
+            elif ext == "docx":
+                template_file_type = "docx"
+
+        # 更新进度 - 准备开始填写
+        await update_task_status(
+            task_id, "processing",
+            progress=10, message=f"准备填写 {len(fields)} 个字段..."
+        )
+
         # 执行填写
         result = await template_fill_service.fill_template(
             template_fields=fields,
             source_doc_ids=request.source_doc_ids,
             source_file_paths=request.source_file_paths,
-            user_hint=request.user_hint
+            user_hint=request.user_hint,
+            template_id=request.template_id,
+            template_file_type=template_file_type,
+            task_id=task_id
         )
 
-        return result
+        # 更新为成功
+        await update_task_status(
+            task_id, "success",
+            progress=100, message="填表完成",
+            result={
+                "field_count": len(fields),
+                "max_rows": result.get("max_rows", 0)
+            }
+        )
+
+        return {**result, "task_id": task_id}
 
     except Exception as e:
+        # 更新为失败
+        await update_task_status(
+            task_id, "failure",
+            progress=0, message="填表失败",
+            error=str(e)
+        )
         logger.error(f"填写表格失败: {str(e)}")
         raise HTTPException(status_code=500, detail=f"填写失败: {str(e)}")
 

backend/app/api/endpoints/upload.py

@@ -5,6 +5,7 @@ from fastapi import APIRouter, UploadFile, File, HTTPException, Query
 from fastapi.responses import StreamingResponse
 from typing import Optional
 import logging
+import os
 import pandas as pd
 import io
 
@@ -126,7 +127,7 @@ async def upload_excel(
                                 content += f"... (共 {len(sheet_data['rows'])} 行)\n\n"
 
             doc_metadata = {
-                "filename": saved_path.split("/")[-1] if "/" in saved_path else saved_path.split("\\")[-1],
+                "filename": os.path.basename(saved_path),
                 "original_filename": file.filename,
                 "saved_path": saved_path,
                 "file_size": len(content),
@@ -253,7 +254,7 @@ async def export_excel(
         output.seek(0)
 
         # 生成文件名
-        original_name = file_path.split('/')[-1] if '/' in file_path else file_path
+        original_name = os.path.basename(file_path)
         if columns:
             export_name = f"export_{sheet_name or 'data'}_{len(column_list) if columns else 'all'}_cols.xlsx"
         else:

backend/app/core/database/mongodb.py

@@ -59,6 +59,11 @@ class MongoDB:
         """RAG索引集合 - 存储字段语义索引"""
         return self.db["rag_index"]
 
+    @property
+    def tasks(self):
+        """任务集合 - 存储任务历史记录"""
+        return self.db["tasks"]
+
     # ==================== 文档操作 ====================
 
     async def insert_document(
@@ -242,8 +247,128 @@ class MongoDB:
         await self.rag_index.create_index("table_name")
         await self.rag_index.create_index("field_name")
 
+        # 任务集合索引
+        await self.tasks.create_index("task_id", unique=True)
+        await self.tasks.create_index("created_at")
+
         logger.info("MongoDB 索引创建完成")
 
+    # ==================== 任务历史操作 ====================
+
+    async def insert_task(
+        self,
+        task_id: str,
+        task_type: str,
+        status: str = "pending",
+        message: str = "",
+        result: Optional[Dict[str, Any]] = None,
+        error: Optional[str] = None,
+    ) -> str:
+        """
+        插入任务记录
+
+        Args:
+            task_id: 任务ID
+            task_type: 任务类型
+            status: 任务状态
+            message: 任务消息
+            result: 任务结果
+            error: 错误信息
+
+        Returns:
+            插入文档的ID
+        """
+        task = {
+            "task_id": task_id,
+            "task_type": task_type,
+            "status": status,
+            "message": message,
+            "result": result,
+            "error": error,
+            "created_at": datetime.utcnow(),
+            "updated_at": datetime.utcnow(),
+        }
+        result_obj = await self.tasks.insert_one(task)
+        return str(result_obj.inserted_id)
+
+    async def update_task(
+        self,
+        task_id: str,
+        status: Optional[str] = None,
+        message: Optional[str] = None,
+        result: Optional[Dict[str, Any]] = None,
+        error: Optional[str] = None,
+    ) -> bool:
+        """
+        更新任务状态
+
+        Args:
+            task_id: 任务ID
+            status: 任务状态
+            message: 任务消息
+            result: 任务结果
+            error: 错误信息
+
+        Returns:
+            是否更新成功
+        """
+        from bson import ObjectId
+
+        update_data = {"updated_at": datetime.utcnow()}
+        if status is not None:
+            update_data["status"] = status
+        if message is not None:
+            update_data["message"] = message
+        if result is not None:
+            update_data["result"] = result
+        if error is not None:
+            update_data["error"] = error
+
+        update_result = await self.tasks.update_one(
+            {"task_id": task_id},
+            {"$set": update_data}
+        )
+        return update_result.modified_count > 0
+
+    async def get_task(self, task_id: str) -> Optional[Dict[str, Any]]:
+        """根据task_id获取任务"""
+        task = await self.tasks.find_one({"task_id": task_id})
+        if task:
+            task["_id"] = str(task["_id"])
+        return task
+
+    async def list_tasks(
+        self,
+        limit: int = 50,
+        skip: int = 0,
+    ) -> List[Dict[str, Any]]:
+        """
+        获取任务列表
+
+        Args:
+            limit: 返回数量
+            skip: 跳过数量
+
+        Returns:
+            任务列表
+        """
+        cursor = self.tasks.find().sort("created_at", -1).skip(skip).limit(limit)
+        tasks = []
+        async for task in cursor:
+            task["_id"] = str(task["_id"])
+            # 转换 datetime 为字符串
+            if task.get("created_at"):
+                task["created_at"] = task["created_at"].isoformat()
+            if task.get("updated_at"):
+                task["updated_at"] = task["updated_at"].isoformat()
+            tasks.append(task)
+        return tasks
+
+    async def delete_task(self, task_id: str) -> bool:
+        """删除任务"""
+        result = await self.tasks.delete_one({"task_id": task_id})
+        return result.deleted_count > 0
+
 
 # ==================== 全局单例 ====================
 

backend/app/core/document_parser/docx_parser.py

@@ -59,7 +59,13 @@ class DocxParser(BaseParser):
             paragraphs = []
             for para in doc.paragraphs:
                 if para.text.strip():
-                    paragraphs.append(para.text)
+                    paragraphs.append({
+                        "text": para.text,
+                        "style": str(para.style.name) if para.style else "Normal"
+                    })
+
+            # 提取段落纯文本（用于 AI 解析）
+            paragraphs_text = [p["text"] for p in paragraphs if p["text"].strip()]
 
             # 提取表格内容
             tables_data = []
@@ -77,8 +83,25 @@ class DocxParser(BaseParser):
                         "column_count": len(table_rows[0]) if table_rows else 0
                     })
 
-            # 合并所有文本
-            full_text = "\n".join(paragraphs)
+            # 提取图片/嵌入式对象信息
+            images_info = self._extract_images_info(doc, path)
+
+            # 合并所有文本（包括图片描述）
+            full_text_parts = []
+            full_text_parts.append("【文档正文】")
+            full_text_parts.extend(paragraphs_text)
+
+            if tables_data:
+                full_text_parts.append("\n【文档表格】")
+                for idx, table in enumerate(tables_data):
+                    full_text_parts.append(f"--- 表格 {idx + 1} ---")
+                    for row in table["rows"]:
+                        full_text_parts.append(" | ".join(str(cell) for cell in row))
+
+            if images_info.get("image_count", 0) > 0:
+                full_text_parts.append(f"\n【文档图片】文档包含 {images_info['image_count']} 张图片/图表")
+
+            full_text = "\n".join(full_text_parts)
 
             # 构建元数据
             metadata = {
@@ -89,7 +112,9 @@ class DocxParser(BaseParser):
                 "table_count": len(tables_data),
                 "word_count": len(full_text),
                 "char_count": len(full_text.replace("\n", "")),
-                "has_tables": len(tables_data) > 0
+                "has_tables": len(tables_data) > 0,
+                "has_images": images_info.get("image_count", 0) > 0,
+                "image_count": images_info.get("image_count", 0)
             }
 
             # 返回结果
@@ -97,12 +122,16 @@ class DocxParser(BaseParser):
                 success=True,
                 data={
                     "content": full_text,
-                    "paragraphs": paragraphs,
+                    "paragraphs": paragraphs_text,
+                    "paragraphs_with_style": paragraphs,
                     "tables": tables_data,
+                    "images": images_info,
                     "word_count": len(full_text),
                     "structured_data": {
                         "paragraphs": paragraphs,
-                        "tables": tables_data
+                        "paragraphs_text": paragraphs_text,
+                        "tables": tables_data,
+                        "images": images_info
                     }
                 },
                 metadata=metadata
@@ -115,6 +144,59 @@ class DocxParser(BaseParser):
                 error=f"解析 Word 文档失败: {str(e)}"
             )
 
+    def extract_images_as_base64(self, file_path: str) -> List[Dict[str, str]]:
+        """
+        提取 Word 文档中的所有图片，返回 base64 编码列表
+
+        Args:
+            file_path: Word 文件路径
+
+        Returns:
+            图片列表，每项包含 base64 编码和图片类型
+        """
+        import zipfile
+        import base64
+        from io import BytesIO
+
+        images = []
+
+        try:
+            with zipfile.ZipFile(file_path, 'r') as zf:
+                # 查找 word/media 目录下的图片文件
+                for filename in zf.namelist():
+                    if filename.startswith('word/media/'):
+                        # 获取图片类型
+                        ext = filename.split('.')[-1].lower()
+                        mime_types = {
+                            'png': 'image/png',
+                            'jpg': 'image/jpeg',
+                            'jpeg': 'image/jpeg',
+                            'gif': 'image/gif',
+                            'bmp': 'image/bmp'
+                        }
+                        mime_type = mime_types.get(ext, 'image/png')
+
+                        try:
+                            # 读取图片数据并转为 base64
+                            image_data = zf.read(filename)
+                            base64_data = base64.b64encode(image_data).decode('utf-8')
+
+                            images.append({
+                                "filename": filename,
+                                "mime_type": mime_type,
+                                "base64": base64_data,
+                                "size": len(image_data)
+                            })
+                            logger.info(f"提取图片: {filename}, 大小: {len(image_data)} bytes")
+                        except Exception as e:
+                            logger.warning(f"提取图片失败 {filename}: {str(e)}")
+
+        except Exception as e:
+            logger.error(f"打开 Word 文档提取图片失败: {str(e)}")
+
+        logger.info(f"共提取 {len(images)} 张图片")
+        return images
+
     def extract_key_sentences(self, text: str, max_sentences: int = 10) -> List[str]:
         """
         从文本中提取关键句子
@@ -268,6 +350,60 @@ class DocxParser(BaseParser):
 
         return fields
 
+    def _extract_images_info(self, doc: Document, path: Path) -> Dict[str, Any]:
+        """
+        提取 Word 文档中的图片/嵌入式对象信息
+
+        Args:
+            doc: Document 对象
+            path: 文件路径
+
+        Returns:
+            图片信息字典
+        """
+        import zipfile
+        from io import BytesIO
+
+        image_count = 0
+        image_descriptions = []
+        inline_shapes_count = 0
+
+        try:
+            # 方法1: 通过 inline shapes 统计图片
+            try:
+                inline_shapes_count = len(doc.inline_shapes)
+                if inline_shapes_count > 0:
+                    image_count = inline_shapes_count
+                    image_descriptions.append(f"文档包含 {inline_shapes_count} 个嵌入式图形/图片")
+            except Exception:
+                pass
+
+            # 方法2: 通过 ZIP 分析 document.xml 获取图片引用
+            try:
+                with zipfile.ZipFile(path, 'r') as zf:
+                    # 查找 word/media 目录下的图片文件
+                    media_files = [f for f in zf.namelist() if f.startswith('word/media/')]
+                    if media_files and not inline_shapes_count:
+                        image_count = len(media_files)
+                        image_descriptions.append(f"文档包含 {image_count} 个嵌入图片")
+
+                    # 检查是否有页眉页脚中的图片
+                    header_images = [f for f in zf.namelist() if 'header' in f.lower() and f.endswith(('.png', '.jpg', '.jpeg', '.gif', '.bmp'))]
+                    if header_images:
+                        image_descriptions.append(f"页眉/页脚包含 {len(header_images)} 个图片")
+            except Exception:
+                pass
+
+        except Exception as e:
+            logger.warning(f"提取图片信息失败: {str(e)}")
+
+        return {
+            "image_count": image_count,
+            "inline_shapes_count": inline_shapes_count,
+            "descriptions": image_descriptions,
+            "has_images": image_count > 0
+        }
+
     def _infer_field_type_from_hint(self, hint: str) -> str:
         """
         从提示词推断字段类型

backend/app/instruction/__init__.py

@@ -0,0 +1,15 @@
+"""
+指令执行模块
+
+注意: 此模块为可选功能，当前尚未实现。
+如需启用，请实现 intent_parser.py 和 executor.py
+"""
+from .intent_parser import IntentParser, DefaultIntentParser
+from .executor import InstructionExecutor, DefaultInstructionExecutor
+
+__all__ = [
+    "IntentParser",
+    "DefaultIntentParser",
+    "InstructionExecutor",
+    "DefaultInstructionExecutor",
+]

backend/app/instruction/executor.py

@@ -0,0 +1,35 @@
+"""
+指令执行器模块
+
+将自然语言指令转换为可执行操作
+
+注意: 此模块为可选功能，当前尚未实现。
+"""
+from abc import ABC, abstractmethod
+from typing import Any, Dict
+
+
+class InstructionExecutor(ABC):
+    """指令执行器抽象基类"""
+
+    @abstractmethod
+    async def execute(self, instruction: str, context: Dict[str, Any]) -> Dict[str, Any]:
+        """
+        执行指令
+
+        Args:
+            instruction: 解析后的指令
+            context: 执行上下文
+
+        Returns:
+            执行结果
+        """
+        pass
+
+
+class DefaultInstructionExecutor(InstructionExecutor):
+    """默认指令执行器"""
+
+    async def execute(self, instruction: str, context: Dict[str, Any]) -> Dict[str, Any]:
+        """暂未实现"""
+        raise NotImplementedError("指令执行功能暂未实现")

backend/app/instruction/intent_parser.py

@@ -0,0 +1,34 @@
+"""
+意图解析器模块
+
+解析用户自然语言指令，识别意图和参数
+
+注意: 此模块为可选功能，当前尚未实现。
+"""
+from abc import ABC, abstractmethod
+from typing import Any, Dict, Tuple
+
+
+class IntentParser(ABC):
+    """意图解析器抽象基类"""
+
+    @abstractmethod
+    async def parse(self, text: str) -> Tuple[str, Dict[str, Any]]:
+        """
+        解析自然语言指令
+
+        Args:
+            text: 用户输入的自然语言
+
+        Returns:
+            (意图类型, 参数字典)
+        """
+        pass
+
+
+class DefaultIntentParser(IntentParser):
+    """默认意图解析器"""
+
+    async def parse(self, text: str) -> Tuple[str, Dict[str, Any]]:
+        """暂未实现"""
+        raise NotImplementedError("意图解析功能暂未实现")

backend/app/main.py

@@ -1,6 +1,13 @@
 """
 FastAPI 应用主入口
 """
+# ========== 压制 MongoDB 疯狂刷屏日志 ==========
+import logging
+logging.getLogger("pymongo").setLevel(logging.WARNING)
+logging.getLogger("pymongo.topology").setLevel(logging.WARNING)
+logging.getLogger("urllib3").setLevel(logging.WARNING)
+# ==============================================
+
 import logging
 import logging.handlers
 import sys

backend/app/services/llm_service.py

@@ -65,7 +65,17 @@ class LLMService:
                 return response.json()
 
         except httpx.HTTPStatusError as e:
-            logger.error(f"LLM API 请求失败: {e.response.status_code} - {e.response.text}")
+            error_detail = e.response.text
+            logger.error(f"LLM API 请求失败: {e.response.status_code} - {error_detail}")
+            # 尝试解析错误信息
+            try:
+                import json
+                err_json = json.loads(error_detail)
+                err_code = err_json.get("error", {}).get("code", "unknown")
+                err_msg = err_json.get("error", {}).get("message", "unknown")
+                logger.error(f"API 错误码: {err_code}, 错误信息: {err_msg}")
+            except:
+                pass
             raise
         except Exception as e:
             logger.error(f"LLM API 调用异常: {str(e)}")
@@ -328,6 +338,154 @@ Excel 数据概览:
                 "analysis": None
             }
 
+    async def chat_with_images(
+        self,
+        text: str,
+        images: List[Dict[str, str]],
+        temperature: float = 0.7,
+        max_tokens: Optional[int] = None
+    ) -> Dict[str, Any]:
+        """
+        调用视觉模型 API（支持图片输入）
+
+        Args:
+            text: 文本内容
+            images: 图片列表，每项包含 base64 编码和 mime_type
+                   格式: [{"base64": "...", "mime_type": "image/png"}, ...]
+            temperature: 温度参数
+            max_tokens: 最大 token 数
+
+        Returns:
+            Dict[str, Any]: API 响应结果
+        """
+        headers = {
+            "Authorization": f"Bearer {self.api_key}",
+            "Content-Type": "application/json"
+        }
+
+        # 构建图片内容
+        image_contents = []
+        for img in images:
+            image_contents.append({
+                "type": "image_url",
+                "image_url": {
+                    "url": f"data:{img['mime_type']};base64,{img['base64']}"
+                }
+            })
+
+        # 构建消息
+        messages = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "text",
+                        "text": text
+                    },
+                    *image_contents
+                ]
+            }
+        ]
+
+        payload = {
+            "model": self.model_name,
+            "messages": messages,
+            "temperature": temperature
+        }
+
+        if max_tokens:
+            payload["max_tokens"] = max_tokens
+
+        try:
+            async with httpx.AsyncClient(timeout=120.0) as client:
+                response = await client.post(
+                    f"{self.base_url}/chat/completions",
+                    headers=headers,
+                    json=payload
+                )
+                response.raise_for_status()
+                return response.json()
+
+        except httpx.HTTPStatusError as e:
+            error_detail = e.response.text
+            logger.error(f"视觉模型 API 请求失败: {e.response.status_code} - {error_detail}")
+            # 尝试解析错误信息
+            try:
+                import json
+                err_json = json.loads(error_detail)
+                err_code = err_json.get("error", {}).get("code", "unknown")
+                err_msg = err_json.get("error", {}).get("message", "unknown")
+                logger.error(f"API 错误码: {err_code}, 错误信息: {err_msg}")
+                logger.error(f"请求模型: {self.model_name}, base_url: {self.base_url}")
+            except:
+                pass
+            raise
+        except Exception as e:
+            logger.error(f"视觉模型 API 调用异常: {str(e)}")
+            raise
+
+    async def analyze_images(
+        self,
+        images: List[Dict[str, str]],
+        user_prompt: str = ""
+    ) -> Dict[str, Any]:
+        """
+        分析图片内容（使用视觉模型）
+
+        Args:
+            images: 图片列表，每项包含 base64 编码和 mime_type
+            user_prompt: 用户提示词
+
+        Returns:
+            Dict[str, Any]: 分析结果
+        """
+        prompt = f"""你是一个专业的视觉分析专家。请分析以下图片内容。
+
+{user_prompt if user_prompt else "请详细描述图片中的内容，包括文字、数据、图表、流程等所有可见信息。"}
+
+请按照以下 JSON 格式输出：
+{{
+    "description": "图片内容的详细描述",
+    "text_content": "图片中的文字内容（如有）",
+    "data_extracted": {{"键": "值"}}  // 如果图片中有表格或数据
+}}
+
+如果图片不包含有用信息，请返回空的描述。"""
+
+        try:
+            response = await self.chat_with_images(
+                text=prompt,
+                images=images,
+                temperature=0.1,
+                max_tokens=4000
+            )
+
+            content = self.extract_message_content(response)
+
+            # 解析 JSON
+            import json
+            try:
+                result = json.loads(content)
+                return {
+                    "success": True,
+                    "analysis": result,
+                    "model": self.model_name
+                }
+            except json.JSONDecodeError:
+                return {
+                    "success": True,
+                    "analysis": {"description": content},
+                    "model": self.model_name
+                }
+
+        except Exception as e:
+            logger.error(f"图片分析失败: {str(e)}")
+            return {
+                "success": False,
+                "error": str(e),
+                "analysis": None
+            }
+
 
 # 全局单例
 llm_service = LLMService()

backend/app/services/rag_service.py

@@ -3,7 +3,6 @@ RAG 服务模块 - 检索增强生成
 
 使用 sentence-transformers + Faiss 实现向量检索
 """
-import json
 import logging
 import os
 import pickle
@@ -11,12 +10,20 @@ from typing import Any, Dict, List, Optional
 
 import faiss
 import numpy as np
-from sentence_transformers import SentenceTransformer
 
 from app.config import settings
 
 logger = logging.getLogger(__name__)
 
+# 尝试导入 sentence-transformers
+try:
+    from sentence_transformers import SentenceTransformer
+    SENTENCE_TRANSFORMERS_AVAILABLE = True
+except ImportError as e:
+    logger.warning(f"sentence-transformers 导入失败: {e}")
+    SENTENCE_TRANSFORMERS_AVAILABLE = False
+    SentenceTransformer = None
+
 
 class SimpleDocument:
     """简化文档对象"""
@@ -28,17 +35,24 @@ class SimpleDocument:
 class RAGService:
     """RAG 检索增强服务"""
 
+    # 默认分块参数
+    DEFAULT_CHUNK_SIZE = 500  # 每个文本块的大小（字符数）
+    DEFAULT_CHUNK_OVERLAP = 50  # 块之间的重叠（字符数）
+
     def __init__(self):
-        self.embedding_model: Optional[SentenceTransformer] = None
+        self.embedding_model = None
         self.index: Optional[faiss.Index] = None
         self.documents: List[Dict[str, Any]] = []
         self.doc_ids: List[str] = []
-        self._dimension: int = 0
+        self._dimension: int = 384  # 默认维度
         self._initialized = False
         self._persist_dir = settings.FAISS_INDEX_DIR
-        # 临时禁用 RAG API 调用，仅记录日志
-        self._disabled = True
-        logger.info("RAG 服务已禁用（_disabled=True），仅记录索引操作日志")
+        # 检查是否可用
+        self._disabled = not SENTENCE_TRANSFORMERS_AVAILABLE
+        if self._disabled:
+            logger.warning("RAG 服务已禁用（sentence-transformers 不可用），将使用关键词匹配作为后备")
+        else:
+            logger.info("RAG 服务已启用")
 
     def _init_embeddings(self):
         """初始化嵌入模型"""
@@ -88,6 +102,63 @@ class RAGService:
         norms = np.where(norms == 0, 1, norms)
         return vectors / norms
 
+    def _split_into_chunks(self, text: str, chunk_size: int = None, overlap: int = None) -> List[str]:
+        """
+        将长文本分割成块
+
+        Args:
+            text: 待分割的文本
+            chunk_size: 每个块的大小（字符数）
+            overlap: 块之间的重叠字符数
+
+        Returns:
+            文本块列表
+        """
+        if chunk_size is None:
+            chunk_size = self.DEFAULT_CHUNK_SIZE
+        if overlap is None:
+            overlap = self.DEFAULT_CHUNK_OVERLAP
+
+        if len(text) <= chunk_size:
+            return [text] if text.strip() else []
+
+        chunks = []
+        start = 0
+        text_len = len(text)
+
+        while start < text_len:
+            # 计算当前块的结束位置
+            end = start + chunk_size
+
+            # 如果不是最后一块，尝试在句子边界处切割
+            if end < text_len:
+                # 向前查找最后一个句号、逗号、换行或分号
+                cut_positions = []
+                for i in range(end, max(start, end - 100), -1):
+                    if text[i] in '。；，,\n、':
+                        cut_positions.append(i + 1)
+                        break
+
+                if cut_positions:
+                    end = cut_positions[0]
+                else:
+                    # 如果没找到句子边界，尝试向后查找
+                    for i in range(end, min(text_len, end + 50)):
+                        if text[i] in '。；，,\n、':
+                            end = i + 1
+                            break
+
+            chunk = text[start:end].strip()
+            if chunk:
+                chunks.append(chunk)
+
+            # 移动起始位置（考虑重叠）
+            start = end - overlap
+            if start <= 0:
+                start = end
+
+        return chunks
+
     def index_field(
         self,
         table_name: str,
@@ -124,9 +195,20 @@ class RAGService:
         self,
         doc_id: str,
         content: str,
-        metadata: Optional[Dict[str, Any]] = None
+        metadata: Optional[Dict[str, Any]] = None,
+        chunk_size: int = None,
+        chunk_overlap: int = None
     ):
-        """将文档内容索引到向量数据库"""
+        """
+        将文档内容索引到向量数据库（自动分块）
+
+        Args:
+            doc_id: 文档唯一标识
+            content: 文档内容
+            metadata: 文档元数据
+            chunk_size: 文本块大小（字符数），默认500
+            chunk_overlap: 块之间的重叠字符数，默认50
+        """
         if self._disabled:
             logger.info(f"[RAG DISABLED] 文档索引操作已跳过: {doc_id}")
             return
@@ -139,18 +221,56 @@ class RAGService:
             logger.debug(f"文档跳过索引 (无嵌入模型): {doc_id}")
             return
 
-        doc = SimpleDocument(
-            page_content=content,
-            metadata=metadata or {"doc_id": doc_id}
-        )
-        self._add_documents([doc], [doc_id])
-        logger.debug(f"已索引文档: {doc_id}")
+        # 分割文档为小块
+        if chunk_size is None:
+            chunk_size = self.DEFAULT_CHUNK_SIZE
+        if chunk_overlap is None:
+            chunk_overlap = self.DEFAULT_CHUNK_OVERLAP
+
+        chunks = self._split_into_chunks(content, chunk_size, chunk_overlap)
+
+        if not chunks:
+            logger.warning(f"文档内容为空，跳过索引: {doc_id}")
+            return
+
+        # 为每个块创建文档对象
+        documents = []
+        chunk_ids = []
+
+        for i, chunk in enumerate(chunks):
+            chunk_id = f"{doc_id}_chunk_{i}"
+            chunk_metadata = metadata.copy() if metadata else {}
+            chunk_metadata.update({
+                "chunk_index": i,
+                "total_chunks": len(chunks),
+                "doc_id": doc_id
+            })
+
+            documents.append(SimpleDocument(
+                page_content=chunk,
+                metadata=chunk_metadata
+            ))
+            chunk_ids.append(chunk_id)
+
+        # 批量添加文档
+        self._add_documents(documents, chunk_ids)
+        logger.info(f"已索引文档 {doc_id}，共 {len(chunks)} 个块")
 
     def _add_documents(self, documents: List[SimpleDocument], doc_ids: List[str]):
         """批量添加文档到向量索引"""
         if not documents:
             return
 
+        # 总是将文档存储在内存中（用于关键词搜索后备）
+        for doc, did in zip(documents, doc_ids):
+            self.documents.append({"id": did, "content": doc.page_content, "metadata": doc.metadata})
+            self.doc_ids.append(did)
+
+        # 如果没有嵌入模型，跳过向量索引
+        if self.embedding_model is None:
+            logger.debug(f"文档跳过向量索引 (无嵌入模型): {len(documents)} 个文档")
+            return
+
         texts = [doc.page_content for doc in documents]
         embeddings = self.embedding_model.encode(texts, convert_to_numpy=True)
         embeddings = self._normalize_vectors(embeddings).astype('float32')
@@ -162,12 +282,18 @@ class RAGService:
         id_array = np.array(id_list, dtype='int64')
         self.index.add_with_ids(embeddings, id_array)
 
-        for doc, did in zip(documents, doc_ids):
-            self.documents.append({"id": did, "content": doc.page_content, "metadata": doc.metadata})
-            self.doc_ids.append(did)
+    def retrieve(self, query: str, top_k: int = 5, min_score: float = 0.3) -> List[Dict[str, Any]]:
+        """
+        根据查询检索相关文档块
 
-    def retrieve(self, query: str, top_k: int = 5) -> List[Dict[str, Any]]:
-        """根据查询检索相关文档"""
+        Args:
+            query: 查询文本
+            top_k: 返回的最大结果数
+            min_score: 最低相似度分数阈值
+
+        Returns:
+            相关文档块列表，每项包含 content, metadata, score, doc_id, chunk_index
+        """
         if self._disabled:
             logger.info(f"[RAG DISABLED] 检索操作已跳过: query={query}, top_k={top_k}")
             return []
@@ -175,9 +301,9 @@ class RAGService:
         if not self._initialized:
             self._init_vector_store()
 
-        if self.index is None or self.index.ntotal == 0:
-            return []
-
+        # 优先使用向量检索
+        if self.index is not None and self.index.ntotal > 0 and self.embedding_model is not None:
+            try:
                 query_embedding = self.embedding_model.encode([query], convert_to_numpy=True)
                 query_embedding = self._normalize_vectors(query_embedding).astype('float32')
 
@@ -187,16 +313,101 @@ class RAGService:
                 for score, idx in zip(scores[0], indices[0]):
                     if idx < 0:
                         continue
+                    if score < min_score:
+                        continue
                     doc = self.documents[idx]
                     results.append({
                         "content": doc["content"],
                         "metadata": doc["metadata"],
                         "score": float(score),
-                "doc_id": doc["id"]
+                        "doc_id": doc["id"],
+                        "chunk_index": doc["metadata"].get("chunk_index", 0)
                     })
 
-        logger.debug(f"检索到 {len(results)} 条相关文档")
+                if results:
+                    logger.debug(f"向量检索到 {len(results)} 条相关文档块")
                     return results
+            except Exception as e:
+                logger.warning(f"向量检索失败，使用关键词搜索后备: {e}")
+
+        # 后备：使用关键词搜索
+        logger.debug("使用关键词搜索后备方案")
+        return self._keyword_search(query, top_k)
+
+    def _keyword_search(self, query: str, top_k: int = 5) -> List[Dict[str, Any]]:
+        """
+        关键词搜索后备方案
+
+        Args:
+            query: 查询文本
+            top_k: 返回的最大结果数
+
+        Returns:
+            相关文档块列表
+        """
+        if not self.documents:
+            return []
+
+        # 提取查询关键词
+        keywords = []
+        for char in query:
+            if '\u4e00' <= char <= '\u9fff':  # 中文字符
+                keywords.append(char)
+        # 添加英文单词
+        import re
+        english_words = re.findall(r'[a-zA-Z]+', query)
+        keywords.extend(english_words)
+
+        if not keywords:
+            return []
+
+        results = []
+        for doc in self.documents:
+            content = doc["content"]
+            # 计算关键词匹配分数
+            score = 0
+            matched_keywords = 0
+            for kw in keywords:
+                if kw in content:
+                    score += 1
+                    matched_keywords += 1
+
+            if matched_keywords > 0:
+                # 归一化分数
+                score = score / max(len(keywords), 1)
+                results.append({
+                    "content": content,
+                    "metadata": doc["metadata"],
+                    "score": score,
+                    "doc_id": doc["id"],
+                    "chunk_index": doc["metadata"].get("chunk_index", 0)
+                })
+
+        # 按分数排序
+        results.sort(key=lambda x: x["score"], reverse=True)
+
+        logger.debug(f"关键词搜索返回 {len(results[:top_k])} 条结果")
+        return results[:top_k]
+
+    def retrieve_by_doc_id(self, doc_id: str, top_k: int = 10) -> List[Dict[str, Any]]:
+        """
+        获取指定文档的所有块
+
+        Args:
+            doc_id: 文档ID
+            top_k: 返回的最大结果数
+
+        Returns:
+            该文档的所有块
+        """
+        # 获取属于该文档的所有块
+        doc_chunks = [d for d in self.documents if d["metadata"].get("doc_id") == doc_id]
+
+        # 按 chunk_index 排序
+        doc_chunks.sort(key=lambda x: x["metadata"].get("chunk_index", 0))
+
+        # 返回指定数量
+        return doc_chunks[:top_k]
 
     def retrieve_by_table(self, table_name: str, top_k: int = 5) -> List[Dict[str, Any]]:
         """检索指定表的字段"""

backend/app/services/template_fill_service.py

@@ -3,6 +3,7 @@
 
 从非结构化文档中检索信息并填写到表格模板
 """
+import asyncio
 import logging
 from dataclasses import dataclass, field
 from typing import Any, Dict, List, Optional
@@ -11,6 +12,7 @@ from app.core.database import mongodb
 from app.services.llm_service import llm_service
 from app.core.document_parser import ParserFactory
 from app.services.markdown_ai_service import markdown_ai_service
+from app.services.rag_service import rag_service
 
 logger = logging.getLogger(__name__)
 
@@ -43,6 +45,7 @@ class FillResult:
     value: Any = ""  # 保留兼容
     source: str = ""  # 来源文档
     confidence: float = 1.0  # 置信度
+    warning: str = None  # 多值提示
 
     def __post_init__(self):
         if self.values is None:
@@ -60,7 +63,10 @@ class TemplateFillService:
         template_fields: List[TemplateField],
         source_doc_ids: Optional[List[str]] = None,
         source_file_paths: Optional[List[str]] = None,
-        user_hint: Optional[str] = None
+        user_hint: Optional[str] = None,
+        template_id: Optional[str] = None,
+        template_file_type: Optional[str] = "xlsx",
+        task_id: Optional[str] = None
     ) -> Dict[str, Any]:
         """
         填写表格模板
@@ -70,6 +76,9 @@ class TemplateFillService:
             source_doc_ids: 源文档 MongoDB ID 列表
             source_file_paths: 源文档文件路径列表
             user_hint: 用户提示（如"请从合同文档中提取"）
+            template_id: 模板文件路径（用于重新生成表头）
+            template_file_type: 模板文件类型
+            task_id: 可选的任务ID，用于任务进度跟踪
 
         Returns:
             填写结果
@@ -94,33 +103,102 @@ class TemplateFillService:
         if not source_docs:
             logger.warning("没有找到源文档，填表结果将全部为空")
 
-        # 2. 对每个字段进行提取
-        for idx, field in enumerate(template_fields):
-            try:
-                logger.info(f"提取字段 [{idx+1}/{len(template_fields)}]: {field.name}")
-                # 从源文档中提取字段值
-                result = await self._extract_field_value(
-                    field=field,
-                    source_docs=source_docs,
-                    user_hint=user_hint
+        # 3. 检查是否需要使用源文档重新生成表头
+        # 条件：源文档已加载 AND 现有字段看起来是自动生成的（如"字段1"、"字段2"）
+        needs_regenerate_headers = (
+            len(source_docs) > 0 and
+            len(template_fields) > 0 and
+            all(self._is_auto_generated_field(f.name) for f in template_fields)
         )
 
-                # 存储结果 - 使用 values 数组
-                filled_data[field.name] = result.values if result.values else [""]
-                fill_details.append({
-                    "field": field.name,
-                    "cell": field.cell,
-                    "values": result.values,
-                    "value": result.value,
-                    "source": result.source,
-                    "confidence": result.confidence
+        if needs_regenerate_headers:
+            logger.info(f"检测到自动生成表头，尝试使用源文档重新生成... (当前字段: {[f.name for f in template_fields]})")
+
+            # 将 SourceDocument 转换为 source_contents 格式
+            source_contents = []
+            for doc in source_docs:
+                structured = doc.structured_data if doc.structured_data else {}
+
+                # 获取标题
+                titles = structured.get("titles", [])
+                if not titles:
+                    titles = []
+
+                # 获取表格
+                tables = structured.get("tables", [])
+                tables_count = len(tables) if tables else 0
+
+                # 生成表格摘要
+                tables_summary = ""
+                if tables:
+                    tables_summary = "\n【文档中的表格】:\n"
+                    for idx, table in enumerate(tables[:5]):
+                        if isinstance(table, dict):
+                            headers = table.get("headers", [])
+                            rows = table.get("rows", [])
+                            if headers:
+                                tables_summary += f"表格{idx+1}表头: {', '.join(str(h) for h in headers)}\n"
+                            if rows:
+                                tables_summary += f"表格{idx+1}前3行: "
+                                for row_idx, row in enumerate(rows[:3]):
+                                    if isinstance(row, list):
+                                        tables_summary += " | ".join(str(c) for c in row) + "; "
+                                    elif isinstance(row, dict):
+                                        tables_summary += " | ".join(str(row.get(h, "")) for h in headers if headers) + "; "
+                                tables_summary += "\n"
+
+                source_contents.append({
+                    "filename": doc.filename,
+                    "doc_type": doc.doc_type,
+                    "content": doc.content[:5000] if doc.content else "",
+                    "titles": titles[:10] if titles else [],
+                    "tables_count": tables_count,
+                    "tables_summary": tables_summary
                 })
 
-                logger.info(f"字段 {field.name} 填写完成: {len(result.values)} 个值")
+            # 使用源文档内容重新生成表头
+            if template_id and template_file_type:
+                logger.info(f"使用源文档重新生成表头: template_id={template_id}, template_file_type={template_file_type}")
+                new_fields = await self.get_template_fields_from_file(
+                    template_id,
+                    template_file_type,
+                    source_contents=source_contents
+                )
+                if new_fields and len(new_fields) > 0:
+                    logger.info(f"成功重新生成表头: {[f.name for f in new_fields]}")
+                    template_fields = new_fields
+                else:
+                    logger.warning("重新生成表头返回空结果，使用原始字段")
+            else:
+                logger.warning("无法重新生成表头：缺少 template_id 或 template_file_type")
+        else:
+            if source_docs and template_fields:
+                logger.info(f"表头看起来正常（非自动生成），无需重新生成: {[f.name for f in template_fields[:5]]}")
 
-            except Exception as e:
-                logger.error(f"填写字段 {field.name} 失败: {str(e)}", exc_info=True)
-                filled_data[field.name] = [f"[提取失败: {str(e)}]"]
+        # 2. 并行提取所有字段（跳过AI审核以提升速度）
+        logger.info(f"开始并行提取 {len(template_fields)} 个字段...")
+
+        # 并行处理所有字段
+        tasks = []
+        for idx, field in enumerate(template_fields):
+            task = self._extract_single_field_fast(
+                field=field,
+                source_docs=source_docs,
+                user_hint=user_hint,
+                field_idx=idx,
+                total_fields=len(template_fields)
+            )
+            tasks.append(task)
+
+        # 等待所有任务完成
+        results = await asyncio.gather(*tasks, return_exceptions=True)
+
+        # 处理结果
+        for idx, result in enumerate(results):
+            field = template_fields[idx]
+            if isinstance(result, Exception):
+                logger.error(f"提取字段 {field.name} 失败: {str(result)}")
+                filled_data[field.name] = [f"[提取失败: {str(result)}]"]
                 fill_details.append({
                     "field": field.name,
                     "cell": field.cell,
@@ -129,6 +207,18 @@ class TemplateFillService:
                     "source": "error",
                     "confidence": 0.0
                 })
+            else:
+                filled_data[field.name] = result.values if result.values else [""]
+                fill_details.append({
+                    "field": field.name,
+                    "cell": field.cell,
+                    "values": result.values,
+                    "value": result.value,
+                    "source": result.source,
+                    "confidence": result.confidence,
+                    "warning": result.warning
+                })
+                logger.info(f"字段 {field.name} 填写完成: {len(result.values) if result.values else 0} 个值")
 
         # 计算最大行数
         max_rows = max(len(v) for v in filled_data.values()) if filled_data else 1
@@ -457,6 +547,353 @@ class TemplateFillService:
                 confidence=0.0
             )
 
+    async def _extract_single_field_fast(
+        self,
+        field: TemplateField,
+        source_docs: List[SourceDocument],
+        user_hint: Optional[str] = None,
+        field_idx: int = 0,
+        total_fields: int = 1
+    ) -> FillResult:
+        """
+        快速提取单个字段（跳过AI审核，减少LLM调用）
+
+        Args:
+            field: 字段定义
+            source_docs: 源文档列表
+            user_hint: 用户提示
+            field_idx: 当前字段索引（用于日志）
+            total_fields: 总字段数（用于日志）
+
+        Returns:
+            提取结果
+        """
+        try:
+            if not source_docs:
+                return FillResult(
+                    field=field.name,
+                    value="",
+                    values=[""],
+                    source="无源文档",
+                    confidence=0.0
+                )
+
+            # 1. 优先尝试直接从结构化数据中提取（最快路径）
+            direct_values = self._extract_values_from_structured_data(source_docs, field.name)
+            if direct_values:
+                logger.info(f"✅ [{field_idx+1}/{total_fields}] 字段 {field.name} 直接从结构化数据提取到 {len(direct_values)} 个值")
+                return FillResult(
+                    field=field.name,
+                    values=direct_values,
+                    value=direct_values[0] if direct_values else "",
+                    source="结构化数据直接提取",
+                    confidence=1.0
+                )
+
+            # 2. 无法直接从结构化数据提取，使用简化版AI提取
+            logger.info(f"🔍 [{field_idx+1}/{total_fields}] 字段 {field.name} 尝试AI提取...")
+
+            # 构建提示词 - 简化版
+            hint_text = field.hint if field.hint else f"请提取{field.name}的信息"
+            if user_hint:
+                hint_text = f"{user_hint}。{hint_text}"
+
+            # 优先使用 RAG 检索内容，否则使用文档开头部分
+            context_parts = []
+            for doc in source_docs:
+                if not doc.content:
+                    logger.info(f"  文档 {doc.filename} 无content内容")
+                    continue
+
+                logger.info(f"  处理文档: {doc.filename}, doc_id={doc.doc_id}, content长度={len(doc.content)}")
+
+                # 尝试 RAG 检索
+                rag_results = rag_service.retrieve(
+                    query=f"{field.name} {hint_text}",
+                    top_k=3,
+                    min_score=0.1
+                )
+
+                if rag_results:
+                    logger.info(f"  RAG检索到 {len(rag_results)} 条结果")
+                    # 使用 RAG 检索到的内容
+                    for r in rag_results:
+                        rag_doc_id = r.get("doc_id", "")
+                        if rag_doc_id.startswith(doc.doc_id):
+                            context_parts.append(r["content"])
+                            logger.info(f"    匹配成功，使用RAG内容，长度={len(r['content'])}")
+                else:
+                    # RAG 没结果，使用文档内容开头
+                    context_parts.append(doc.content[:2500])
+                    logger.info(f"  RAG无结果，使用文档开头 {min(2500, len(doc.content))} 字符")
+
+            context = "\n\n".join(context_parts[:3]) if context_parts else ""
+            logger.info(f"  最终context长度: {len(context)}, 内容预览: {context[:200] if context else '无'}...")
+
+            prompt = f"""你是一个专业的数据提取专家。请严格按照表头字段「{field.name}」从文档中提取数据。
+
+提示: {hint_text}
+
+【重要规则 - 必须遵守】
+1. **每个值必须有标注**：根据数据来源添加合适的标注前缀！
+   - ✅ 正确格式：
+     - "2024年：38710个"
+     - "北京：1234万人次"
+     - "某省：5678万人"
+     - "公立医院：11754个"
+     - "三级医院：4111个"
+     - "图书馆：3246个"
+   - ❌ 错误格式："38710个"（缺少标注）
+
+2. **标注类型根据数据决定**：
+   - 年份类数据 → "2024年：xxx"、"2023年：xxx"
+   - 地区类数据 → "北京：xxx"、"广东：xxx"、"某县：xxx"
+   - 机构/分类数据 → "公立医院：xxx"、"三级医院：xxx"、"图书馆：xxx"
+   - 其他分类 → 根据实际情况标注
+
+3. **严格按表头提取**：只提取与「{field.name}」直接相关的数据
+
+4. **多值必须全部提取并标注**：如果文档中提到多个相关数据，每个都要有标注
+
+文档内容：
+{context if context else "（无文档内容）"}
+
+请严格按格式返回JSON：{{"values": ["标注：数值", "标注：数值", ...]}}
+注意：values数组中每个元素都必须包含标注前缀，不能只有数值！
+"""
+
+            messages = [
+                {"role": "system", "content": "你是一个专业的数据提取助手，擅长从政府统计公报中提取数据。严格按JSON格式输出，只返回values数组。"},
+                {"role": "user", "content": prompt}
+            ]
+
+            response = await self.llm.chat(
+                messages=messages,
+                temperature=0.1,
+                max_tokens=1000
+            )
+
+            content = self.llm.extract_message_content(response)
+            logger.info(f"  LLM原始返回: {content[:500]}")
+
+            # 解析JSON
+            import json
+            import re
+            cleaned = content.strip()
+
+            # 查找JSON开始位置
+            json_start = -1
+            for i, c in enumerate(cleaned):
+                if c == '{':
+                    json_start = i
+                    break
+
+            values = []
+            source = "AI提取"
+            if json_start >= 0:
+                try:
+                    json_text = cleaned[json_start:]
+                    result = json.loads(json_text)
+                    values = result.get("values", [])
+                    logger.info(f"  JSON解析成功，values: {values}")
+                except json.JSONDecodeError as e:
+                    logger.warning(f"  JSON解析失败: {e}，尝试修复...")
+                    # 尝试修复常见JSON问题
+                    try:
+                        # 尝试找到values数组
+                        values_match = re.search(r'"values"\s*:\s*\[(.*?)\]', cleaned, re.DOTALL)
+                        if values_match:
+                            values_str = values_match.group(1)
+                            # 提取数组中的字符串
+                            values = re.findall(r'"([^"]*)"', values_str)
+                            logger.info(f"  正则提取values: {values}")
+                    except:
+                        pass
+
+            # 如果values为空，尝试从文本中用正则提取数字+单位
+            if not values or values == [""]:
+                logger.info(f"  JSON解析未获取到值，尝试正则提取...")
+                # 匹配数字+单位或百分号的模式
+                patterns = [
+                    r'(\d+\.?\d*[亿万千百十个]?[%‰℃℃万元亿]?)',  # 通用数字+单位
+                    r'(\d+\.?\d*%)',  # 百分号
+                    r'(\d+\.?\d*[个万人亿元]?)',  # 中文单位
+                ]
+                for pattern in patterns:
+                    matches = re.findall(pattern, context)
+                    if matches:
+                        # 过滤掉纯数字
+                        filtered = [m for m in matches if not m.replace('.', '').isdigit()]
+                        if filtered:
+                            values = filtered[:10]  # 最多取10个
+                            logger.info(f"  正则提取到: {values}")
+                            break
+
+            if not values or values == [""]:
+                values = self._extract_values_by_regex(cleaned)
+
+            if not values:
+                values = [""]
+
+            # 生成多值提示（基于实际检测到的值数量）
+            warning = ""
+            if len(values) > 1:
+                warning = f"⚠️ 检测到 {len(values)} 个值：{values[:5]}{'...' if len(values) > 5 else ''}"
+
+            logger.info(f"✅ [{field_idx+1}/{total_fields}] 字段 {field.name} AI提取完成: {len(values)} 个值")
+            if warning:
+                logger.info(f"   {warning}")
+
+            return FillResult(
+                field=field.name,
+                values=values,
+                value=values[0] if values else "",
+                source=source,
+                confidence=0.8,
+                warning=warning if warning else None
+            )
+
+        except Exception as e:
+            logger.error(f"❌ [{field_idx+1}/{total_fields}] 字段 {field.name} 提取失败: {str(e)}")
+            return FillResult(
+                field=field.name,
+                values=[""],
+                value="",
+                source=f"提取失败: {str(e)}",
+                confidence=0.0
+            )
+
+    async def _verify_field_value(
+        self,
+        field: TemplateField,
+        extracted_values: List[str],
+        source_docs: List[SourceDocument],
+        user_hint: Optional[str] = None
+    ) -> Optional[FillResult]:
+        """
+        验证并修正提取的字段值
+
+        Args:
+            field: 字段定义
+            extracted_values: 已提取的值
+            source_docs: 源文档列表
+            user_hint: 用户提示
+
+        Returns:
+            验证后的结果，如果验证通过返回None（使用原结果）
+        """
+        if not extracted_values or not extracted_values[0]:
+            return None
+
+        if not source_docs:
+            return None
+
+        try:
+            # 构建验证上下文
+            context_text = self._build_context_text(source_docs, field_name=field.name, max_length=15000)
+
+            hint_text = field.hint if field.hint else f"请理解{field.name}字段的含义"
+            if user_hint:
+                hint_text = f"{user_hint}。{hint_text}"
+
+            prompt = f"""你是一个数据质量审核专家。请审核以下提取的数据是否合理。
+
+【待审核字段】
+字段名：{field.name}
+字段说明：{hint_text}
+
+【已提取的值】
+{extracted_values[:10]}  # 最多审核前10个值
+
+【源文档上下文】
+{context_text[:8000]}
+
+【审核要求】
+1. 这些值是否符合字段的含义？
+2. 值在原文中的原始含义是什么？检查是否有误解或误提取
+3. 是否存在明显错误、空值或不合理的数据？
+4. 如果表格有多个列，请确认提取的是正确的列
+
+请严格按照以下 JSON 格式输出（只需输出 JSON，不要其他内容）：
+{{
+    "is_valid": true或false,
+    "corrected_values": ["修正后的值列表"] 或 null（如果无需修正）,
+    "reason": "审核说明，解释判断理由",
+    "original_meaning": "值在原文中的原始含义描述"
+}}
+"""
+
+            messages = [
+                {"role": "system", "content": "你是一个严格的数据质量审核专家。请仔细核对原文和提取的值是否匹配。"},
+                {"role": "user", "content": prompt}
+            ]
+
+            response = await self.llm.chat(
+                messages=messages,
+                temperature=0.2,
+                max_tokens=3000
+            )
+
+            content = self.llm.extract_message_content(response)
+            logger.info(f"字段 {field.name} 审核返回: {content[:300]}")
+
+            # 解析 JSON
+            import json
+            import re
+
+            cleaned = content.strip()
+            cleaned = re.sub(r'^```json\s*', '', cleaned, flags=re.MULTILINE)
+            cleaned = re.sub(r'^```\s*', '', cleaned, flags=re.MULTILINE)
+            cleaned = cleaned.strip()
+
+            json_start = -1
+            for i, c in enumerate(cleaned):
+                if c == '{':
+                    json_start = i
+                    break
+
+            if json_start == -1:
+                logger.warning(f"字段 {field.name} 审核：无法找到 JSON")
+                return None
+
+            json_text = cleaned[json_start:]
+            result = json.loads(json_text)
+
+            is_valid = result.get("is_valid", True)
+            corrected_values = result.get("corrected_values")
+            reason = result.get("reason", "")
+            original_meaning = result.get("original_meaning", "")
+
+            logger.info(f"字段 {field.name} 审核结果: is_valid={is_valid}, reason={reason[:100]}")
+
+            if not is_valid and corrected_values:
+                # 值有问题且有修正建议，使用修正后的值
+                logger.info(f"字段 {field.name} 使用修正后的值: {corrected_values[:5]}")
+                return FillResult(
+                    field=field.name,
+                    values=corrected_values,
+                    value=corrected_values[0] if corrected_values else "",
+                    source=f"AI审核修正: {reason[:100]}",
+                    confidence=0.7
+                )
+            elif not is_valid and original_meaning:
+                # 值有问题但无修正，记录原始含义供用户参考
+                logger.info(f"字段 {field.name} 审核发现问题: {original_meaning}")
+                return FillResult(
+                    field=field.name,
+                    values=extracted_values,
+                    value=extracted_values[0] if extracted_values else "",
+                    source=f"AI审核疑问: {original_meaning[:100]}",
+                    confidence=0.5
+                )
+
+            # 验证通过，返回 None 表示使用原结果
+            return None
+
+        except Exception as e:
+            logger.error(f"字段 {field.name} 审核失败: {str(e)}")
+            return None
+
     def _build_context_text(self, source_docs: List[SourceDocument], field_name: str = None, max_length: int = 8000) -> str:
         """
         构建上下文文本
@@ -947,13 +1384,148 @@ class TemplateFillService:
 
         values = []
         for row in rows:
-            if isinstance(row, list) and target_idx < len(row):
+            if isinstance(row, list):
+                # 跳过子表头行（主要包含年份值的行，如 "1985", "1995"）
+                if self._is_year_subheader_row(row):
+                    logger.info(f"跳过子表头行: {row[:5]}...")
+                    continue
+                # 跳过章节标题行
+                if self._is_section_header_row(row):
+                    logger.info(f"跳过章节标题行: {row[:5]}...")
+                    continue
+                if target_idx < len(row):
                     val = row[target_idx]
                 else:
                     val = ""
+            else:
+                val = ""
             values.append(self._format_value(val))
 
-        return values
+        # 过滤掉无效值（章节标题、省略号等）
+        valid_values = self._filter_valid_values(values)
+        if len(valid_values) < len(values):
+            logger.info(f"过滤无效值: {len(values)} -> {len(valid_values)}")
+
+        return valid_values
+
+    def _is_year_subheader_row(self, row: List) -> bool:
+        """
+        检测行是否看起来像年份子表头行
+
+        年份子表头行通常包含 "1985", "1995", "2020" 等4位数字
+
+        Args:
+            row: 行数据
+
+        Returns:
+            是否是年份子表头行
+        """
+        if not row:
+            return False
+
+        import re
+        year_pattern = re.compile(r'^(19|20)\d{2}$')
+
+        # 计算看起来像年份的单元格数量
+        year_like_count = 0
+        for cell in row:
+            cell_str = str(cell).strip()
+            if year_pattern.match(cell_str):
+                year_like_count += 1
+
+        # 如果超过50%的单元格是年份格式，认为是子表头行
+        if len(row) > 0 and year_like_count / len(row) > 0.5:
+            return True
+
+        return False
+
+    def _is_section_header_row(self, row: List) -> bool:
+        """
+        检测行是否看起来像章节标题行
+
+        章节标题行通常包含 "其中："、"全部工业中："、"按...计算" 等关键词
+
+        Args:
+            row: 行数据
+
+        Returns:
+            是否是章节标题行
+        """
+        if not row:
+            return False
+
+        import re
+        # 章节标题通常包含这些模式
+        section_patterns = [
+            r'其中[:：]',
+            r'全部\w+中[:：]',
+            r'按\w+计算',
+            r'小计',
+            r'合计',
+            r'总计',
+            r'^其中$',
+            r'全部$'
+        ]
+
+        for cell in row:
+            cell_str = str(cell).strip()
+            if not cell_str:
+                continue
+            for pattern in section_patterns:
+                if re.search(pattern, cell_str):
+                    return True
+
+        return False
+
+    def _is_valid_data_value(self, val: str) -> bool:
+        """
+        检测值是否是有效的数据值（不是章节标题、省略号等）
+
+        Args:
+            val: 值字符串
+
+        Returns:
+            是否是有效数据值
+        """
+        if not val or not str(val).strip():
+            return False
+
+        val_str = str(val).strip()
+
+        # 无效模式
+        invalid_patterns = [
+            r'^…$',  # 省略号
+            r'^[\.。]+$',  # 只有点或句号
+            r'其中[:：]',  # 章节标题
+            r'全部\w+中',  # 章节标题
+            r'按\w+计算',  # 计算类型
+            r'^(小计|合计|总计)$',  # 汇总行
+            r'^其中$',
+            r'^全部$'
+        ]
+
+        for pattern in invalid_patterns:
+            import re
+            if re.match(pattern, val_str):
+                return False
+
+        return True
+
+    def _filter_valid_values(self, values: List[str]) -> List[str]:
+        """
+        过滤出有效的数据值
+
+        Args:
+            values: 值列表
+
+        Returns:
+            只包含有效值的列表
+        """
+        valid_values = []
+        for val in values:
+            if self._is_valid_data_value(val):
+                valid_values.append(val)
+        return valid_values
 
     def _find_best_matching_column(self, headers: List, field_name: str) -> Optional[int]:
         """
@@ -981,6 +1553,11 @@ class TemplateFillService:
             header_str = str(header).strip()
             header_lower = header_str.lower()
 
+            # 跳过空表头（第一列为空的情况）
+            if not header_str:
+                logger.info(f"跳过空表头列: 索引 {idx}")
+                continue
+
             # 策略1: 精确匹配（忽略大小写）
             if header_lower == field_lower:
                 return idx
@@ -1037,6 +1614,12 @@ class TemplateFillService:
 
         values = []
         for row in rows:
+            # 跳过子表头行（主要包含年份值的行，如 "1985", "1995"）
+            if isinstance(row, list) and self._is_year_subheader_row(row):
+                continue
+            # 跳过章节标题行
+            if isinstance(row, list) and self._is_section_header_row(row):
+                continue
             if isinstance(row, dict):
                 val = row.get(target_col, "")
             elif isinstance(row, list) and target_idx < len(row):
@@ -1045,7 +1628,12 @@ class TemplateFillService:
                 val = ""
             values.append(self._format_value(val))
 
-        return values
+        # 过滤掉无效值（章节标题、省略号等）
+        valid_values = self._filter_valid_values(values)
+        if len(valid_values) < len(values):
+            logger.info(f"过滤无效值: {len(values)} -> {len(valid_values)}")
+
+        return valid_values
 
     def _format_value(self, val: Any) -> str:
         """
@@ -1379,38 +1967,98 @@ class TemplateFillService:
                 if user_hint:
                     hint_text = f"{user_hint}。{hint_text}"
 
-                # 构建针对字段提取的提示词
-                prompt = f"""你是一个专业的数据提取专家。请从以下文档内容中提取与"{field.name}"完全匹配的数据。
+                # 构建查询文本
+                query_text = f"{field.name} {hint_text}"
+
+                # 使用 RAG 向量检索获取相关内容块
+                rag_results = rag_service.retrieve(
+                    query=query_text,
+                    top_k=5,
+                    min_score=0.3
+                )
+
+                # 构建上下文：优先使用 RAG 检索结果，如果检索不到则使用原始内容
+                if rag_results:
+                    # 使用 RAG 检索到的相关块
+                    context_parts = []
+                    for result in rag_results:
+                        if result.get("doc_id", "").startswith(doc.doc_id) or not result.get("doc_id"):
+                            context_parts.append(result["content"])
+
+                    if context_parts:
+                        retrieved_context = "\n\n---\n\n".join(context_parts)
+                        logger.info(f"RAG 检索到 {len(context_parts)} 个相关块用于字段 {field.name}")
+                        # 使用检索到的内容（限制长度）
+                        context_to_use = retrieved_context[:6000]
+                    else:
+                        # RAG 检索结果不属于当前文档，使用原始内容
+                        context_to_use = doc.content[:6000] if doc.content else ""
+                        logger.info(f"字段 {field.name} 使用原始内容（RAG结果不属于当前文档）")
+                else:
+                    # 没有 RAG 检索结果，使用原始内容
+                    context_to_use = doc.content[:6000] if doc.content else ""
+                    logger.info(f"字段 {field.name} 使用原始内容（无RAG检索结果）")
+
+                # 构建针对字段提取的提示词 - 增强语义匹配能力
+                prompt = f"""你是一个专业的数据提取专家。请从以下文档内容中进行**语义匹配**提取。
 
 【重要】字段名: "{field.name}"
 【重要】字段提示: {hint_text}
 
-请严格按照以下步骤操作：
-1. 在文档中搜索与"{field.name}"完全相同或高度相关的关键词
-2. 找到后，提取该关键词后的数值（注意：只要数值，不要单位）
-3. 如果是表格中的数据，直接提取该单元格的数值
-4. 如果是段落描述，在关键词附近找数值
+## 分类数据识别
+
+文档中经常包含分类统计数据，格式如下：
+
+### 1. 直接分类（用"其中："、"中，"等分隔）
+原文示例：
+- "全国医疗卫生机构总数1093551个，其中：医院38710个，基层医疗卫生机构1040023个"
+  → 字段"医院数量" 应提取: 38710
+  → 字段"基层医疗卫生机构数量" 应提取: 1040023
+
+- "医院中，公立医院11754个，民营医院26956个"
+  → 字段"公立医院数量" 应提取: 11754
+  → 字段"民营医院数量" 应提取: 26956
+
+### 2. 嵌套分类（用"按...分："、"其中："等结构）
+原文示例：
+- "医院按等级分：三级医院4111个（其中：三级甲等医院1876个），二级医院12294个"
+  → 字段"三级医院数量" 应提取: 4111
+  → 字段"三级甲等医院数量" 应提取: 1876
+  → 字段"二级医院数量" 应提取: 12294
+
+### 3. 匹配技巧
+- "医院数量" 可匹配: "医院38710个"、"医院数量为"
+- "公立医院数量" 可匹配: "公立医院11754个"、"公立医院有"
+- 忽略"数量"、"数"、"个"等后缀的差异
+- 数值可能紧跟关键词，也可能分开描述
+
+## 提取规则
+
+1. **全文搜索**：在文档的全部内容中搜索，不要只搜索开头部分
+2. **分类定位**：找到包含该分类关键词的句子，理解其完整的数值
+3. **保留单位**：提取数值时**要包含单位**
 
 【重要】返回值规则：
-- 只返回纯数值，不要单位（如 "4.9" 而不是 "4.9万亿元"）
-- 如果原文是"4.9万亿元"，返回 "4.9"
-- 如果原文是"144000万册"，返回 "144000"
-- 如果是百分比如"增长7.7%"，返回 "7.7"
-- 如果没有找到完全匹配的数据，返回空数组
+- **返回数值时必须包含单位**
+- 例如原文"公共图书馆3246个"，提取时应返回 "3246个"
+- 例如原文"国内旅游收入4.9万亿元"，提取时应返回 "4.9万亿元"
+- 例如原文"注册护士585.5万人"，提取时应返回 "585.5万人"
+- 如果字段是"指标"类型，返回具体的指标名称文本（不带单位）
+- 如果没有找到任何相关数据，返回空数组
 
 文档内容：
-{doc.content[:10000] if doc.content else ""}
+{context_to_use}
 
 请用严格的 JSON 格式返回：
 {{
-    "values": ["值1", "值2", ...],  // 只填数值，不要单位
-    "source": "数据来源说明",
+    "values": ["提取到的值1", "值2", ...],
+    "source": "数据来源说明（如：从文档第X段提取）",
     "confidence": 0.0到1.0之间的置信度
 }}
 
-示例：
-- 如果字段是"图书馆总藏量（万册）"且文档说"图书总藏量14.4亿册"，返回 values: ["144000"]
-- 如果字段是"国内旅游收入（亿元）"且文档说"国内旅游收入4.9万亿元"，返回 values: ["49000"]"""
+【重要】即使是模糊匹配，也要：
+- 确保提取的内容确实来自文档
+- source中准确说明数据来源位置"""
 
                 messages = [
                     {"role": "system", "content": "你是一个专业的数据提取助手，擅长从政府统计公报等文档中提取数据。请严格按JSON格式输出。"},
@@ -1504,11 +2152,14 @@ class TemplateFillService:
             import pandas as pd
 
             # 读取 Excel 内容检查是否为空
+            content_sample = ""
             if file_type in ["xlsx", "xls"]:
                 df = pd.read_excel(file_path, header=None)
                 if df.shape[0] == 0 or df.shape[1] == 0:
                     logger.info("Excel 表格为空")
-                    # 生成默认字段
+                    # 即使 Excel 为空，如果有源文档，仍然尝试使用 AI 生成表头
+                    if not source_contents:
+                        logger.info("Excel 为空且没有源文档，使用默认字段名")
                         return [TemplateField(
                             cell=self._column_to_cell(i),
                             name=f"字段{i+1}",
@@ -1516,7 +2167,9 @@ class TemplateFillService:
                             required=False,
                             hint="请填写此字段"
                         ) for i in range(5)]
-
+                    # 有源文档，继续调用 AI 生成表头
+                    logger.info("Excel 为空但有源文档，使用源文档内容生成表头...")
+                else:
                     # 表格有数据但没有表头
                     if df.shape[1] > 0:
                         # 读取第一行作为参考，看是否为空
@@ -1532,7 +2185,10 @@ class TemplateFillService:
             # 调用 AI 生成表头
             # 根据源文档内容生成表头
             source_info = ""
+            logger.info(f"[DEBUG] _generate_fields_with_ai received source_contents: {len(source_contents) if source_contents else 0} items")
             if source_contents:
+                for sc in source_contents:
+                    logger.info(f"[DEBUG]   source doc: filename={sc.get('filename')}, content_len={len(sc.get('content', ''))}, titles={len(sc.get('titles', []))}, tables_count={sc.get('tables_count', 0)}, has_tables_summary={bool(sc.get('tables_summary'))}")
                 source_info = "\n\n【源文档内容摘要】（根据以下文档内容生成表头）：\n"
                 for idx, src in enumerate(source_contents[:5]):  # 最多5个源文档
                     filename = src.get("filename", f"文档{idx+1}")
@@ -1540,37 +2196,62 @@ class TemplateFillService:
                     content = src.get("content", "")[:3000]  # 限制内容长度
                     titles = src.get("titles", [])[:10]  # 最多10个标题
                     tables_count = src.get("tables_count", 0)
+                    tables_summary = src.get("tables_summary", "")
 
                     source_info += f"\n--- 文档 {idx+1}: {filename} ({doc_type}) ---\n"
+                    # 处理 titles（可能是字符串列表或字典列表）
                     if titles:
-                        source_info += f"【章节标题】: {', '.join([t.get('text', '') for t in titles[:5]])}\n"
+                        title_texts = []
+                        for t in titles[:5]:
+                            if isinstance(t, dict):
+                                title_texts.append(t.get('text', ''))
+                            else:
+                                title_texts.append(str(t))
+                        if title_texts:
+                            source_info += f"【章节标题】: {', '.join(title_texts)}\n"
                     if tables_count > 0:
                         source_info += f"【包含表格数】: {tables_count}\n"
+                    if tables_summary:
+                        source_info += f"{tables_summary}\n"
                     if content:
-                        source_info += f"【内容预览】: {content[:1500]}...\n"
+                        source_info += f"【文档内容】（前3000字符）：{content[:3000]}\n"
 
-            prompt = f"""你是一个专业的表格设计助手。请根据源文档内容生成合适的表格表头字段。
+            prompt = f"""你是一个专业的数据分析助手。请分析源文档中的所有数据，生成表格表头字段。
 
-任务：用户有一些源文档（可能包含表格数据、统计信息等），需要填写到表格中。请分析源文档内容，生成适合的表头字段。
+任务：分析源文档，找出所有具体的数据指标及其分类。
 
 {source_info}
 
-请生成5-10个简洁的表头字段名，这些字段应该：
-1. 简洁明了，易于理解
-2. 适合作为表格列标题
-3. 直接对应源文档中的关键数据项
-4. 字段之间有明显的区分度
+【重要要求】
+1. **只生成数据字段名**：
+   - ✅ 正确示例："医院数量"、"公立医院数量"、"病床使用率"
+   - ❌ 错误示例："source"、"备注"、"说明"、"数据来源"
+
+2. **识别所有数值数据**：
+   - 例如："医院38710个"、"病床使用率78.8%"
+   - 例如："公立医院11754个"、"公立医院病床使用率84.8%"
+
+3. **理解分类层级**：
+   - 顶级分类：如"医院"、"基层医疗卫生机构"
+   - 二级分类：如"医院"下分为"公立医院"、"民营医院"
+
+4. **生成字段**：
+   - 字段名要简洁，如："医院数量"、"病床使用率"
+   - 优先选择：总数 + 主要分类
+
+5. **生成数量**：
+   - 生成5-7个最有代表性的字段
 
 请严格按照以下 JSON 格式输出（只需输出 JSON，不要其他内容）：
 {{
     "fields": [
-        {{"name": "字段名1", "hint": "字段说明提示1"}},
-        {{"name": "字段名2", "hint": "字段说明提示2"}}
+        {{"name": "字段名1"}},
+        {{"name": "字段名2"}}
     ]
 }}
 """
             messages = [
-                {"role": "system", "content": "你是一个专业的表格设计助手。请严格按JSON格式输出。"},
+                {"role": "system", "content": "你是一个专业的表格设计助手。请严格按JSON格式输出，只返回纯数据字段名，不要source、备注、说明等辅助字段。"},
                 {"role": "user", "content": prompt}
             ]
 
@@ -1609,14 +2290,22 @@ class TemplateFillService:
 
             if result and "fields" in result:
                 fields = []
+                # 过滤非数据字段
+                skip_keywords = ["source", "来源", "备注", "说明", "备注列", "说明列", "data_source", "remark", "note", "description"]
                 for idx, f in enumerate(result["fields"]):
+                    field_name = f.get("name", f"字段{idx+1}")
+                    # 跳过非数据字段
+                    if any(kw in field_name.lower() for kw in skip_keywords):
+                        logger.info(f"跳过非数据字段: {field_name}")
+                        continue
                     fields.append(TemplateField(
                         cell=self._column_to_cell(idx),
-                        name=f.get("name", f"字段{idx+1}"),
+                        name=field_name,
                         field_type="text",
                         required=False,
                         hint=f.get("hint", "")
                     ))
+                logger.info(f"AI 生成表头: {[f.name for f in fields]}")
                 return fields
 
         except Exception as e:

backend/app/services/word_ai_service.py

@@ -0,0 +1,637 @@
+"""
+Word 文档 AI 解析服务
+
+使用 LLM (GLM) 对 Word 文档进行深度理解，提取结构化数据
+"""
+import logging
+from typing import Dict, Any, List, Optional
+import json
+
+from app.services.llm_service import llm_service
+from app.core.document_parser.docx_parser import DocxParser
+
+logger = logging.getLogger(__name__)
+
+
+class WordAIService:
+    """Word 文档 AI 解析服务"""
+
+    def __init__(self):
+        self.llm = llm_service
+        self.parser = DocxParser()
+
+    async def parse_word_with_ai(
+        self,
+        file_path: str,
+        user_hint: str = ""
+    ) -> Dict[str, Any]:
+        """
+        使用 AI 解析 Word 文档，提取结构化数据
+
+        适用于从非结构化的 Word 文档中提取表格数据、键值对等信息
+
+        Args:
+            file_path: Word 文件路径
+            user_hint: 用户提示词，指定要提取的内容类型
+
+        Returns:
+            Dict: 包含结构化数据的解析结果
+        """
+        try:
+            # 1. 先用基础解析器提取原始内容
+            parse_result = self.parser.parse(file_path)
+
+            if not parse_result.success:
+                return {
+                    "success": False,
+                    "error": parse_result.error,
+                    "structured_data": None
+                }
+
+            # 2. 获取原始数据
+            raw_data = parse_result.data
+            paragraphs = raw_data.get("paragraphs", [])
+            paragraphs_with_style = raw_data.get("paragraphs_with_style", [])
+            tables = raw_data.get("tables", [])
+            content = raw_data.get("content", "")
+            images_info = raw_data.get("images", {})
+            metadata = parse_result.metadata or {}
+
+            image_count = images_info.get("image_count", 0)
+            image_descriptions = images_info.get("descriptions", [])
+
+            logger.info(f"Word 基础解析完成: {len(paragraphs)} 个段落, {len(tables)} 个表格, {image_count} 张图片")
+
+            # 3. 提取图片数据（用于视觉分析）
+            images_base64 = []
+            if image_count > 0:
+                try:
+                    images_base64 = self.parser.extract_images_as_base64(file_path)
+                    logger.info(f"提取到 {len(images_base64)} 张图片的 base64 数据")
+                except Exception as e:
+                    logger.warning(f"提取图片 base64 失败: {str(e)}")
+
+            # 4. 根据内容类型选择 AI 解析策略
+            # 如果有图片，先分析图片
+            image_analysis = ""
+            if images_base64:
+                image_analysis = await self._analyze_images_with_ai(images_base64, user_hint)
+                logger.info(f"图片 AI 分析完成: {len(image_analysis)} 字符")
+
+            # 优先处理：表格 > (表格+文本) > 纯文本
+            if tables and len(tables) > 0:
+                structured_data = await self._extract_tables_with_ai(
+                    tables, paragraphs, image_count, user_hint, metadata, image_analysis
+                )
+            elif paragraphs and len(paragraphs) > 0:
+                structured_data = await self._extract_from_text_with_ai(
+                    paragraphs, content, image_count, image_descriptions, user_hint, image_analysis
+                )
+            else:
+                structured_data = {
+                    "success": True,
+                    "type": "empty",
+                    "message": "文档内容为空"
+                }
+
+            # 添加图片分析结果
+            if image_analysis:
+                structured_data["image_analysis"] = image_analysis
+
+            return structured_data
+
+        except Exception as e:
+            logger.error(f"AI 解析 Word 文档失败: {str(e)}")
+            return {
+                "success": False,
+                "error": str(e),
+                "structured_data": None
+            }
+
+    async def _extract_tables_with_ai(
+        self,
+        tables: List[Dict],
+        paragraphs: List[str],
+        image_count: int,
+        user_hint: str,
+        metadata: Dict,
+        image_analysis: str = ""
+    ) -> Dict[str, Any]:
+        """
+        使用 AI 从 Word 表格和文本中提取结构化数据
+
+        Args:
+            tables: 表格列表
+            paragraphs: 段落列表
+            image_count: 图片数量
+            user_hint: 用户提示
+            metadata: 文档元数据
+            image_analysis: 图片 AI 分析结果
+
+        Returns:
+            结构化数据
+        """
+        try:
+            # 构建表格文本描述
+            tables_text = self._build_tables_description(tables)
+
+            # 构建段落描述
+            paragraphs_text = "\n".join(paragraphs[:50]) if paragraphs else "（无正文文本）"
+            if len(paragraphs) > 50:
+                paragraphs_text += f"\n...（共 {len(paragraphs)} 个段落，仅显示前50个）"
+
+            # 图片提示
+            image_hint = f"注意：此文档包含 {image_count} 张图片/图表。" if image_count > 0 else ""
+
+            prompt = f"""你是一个专业的数据提取专家。请从以下 Word 文档的完整内容中提取结构化数据。
+
+【用户需求】
+{user_hint if user_hint else "请提取文档中的所有结构化数据，包括表格数据、键值对、列表项等。"}
+
+【文档正文（段落）】
+{paragraphs_text}
+
+【文档表格】
+{tables_text}
+
+【文档图片信息】
+{image_hint}
+
+请按照以下 JSON 格式输出：
+{{
+    "type": "table_data",
+    "headers": ["列1", "列2", ...],
+    "rows": [["行1列1", "行1列2", ...], ["行2列1", "行2列2", ...], ...],
+    "key_values": {{"键1": "值1", "键2": "值2", ...}},
+    "list_items": ["项1", "项2", ...],
+    "description": "文档内容描述"
+}}
+
+重点：
+- 优先从表格中提取结构化数据
+- 如果表格中有表头，headers 是表头，rows 是数据行
+- 如果文档中有键值对（如 名称: 张三），提取到 key_values 中
+- 如果文档中有列表项，提取到 list_items 中
+- 图片内容无法直接提取，但请在 description 中说明图片的大致主题（如"包含流程图"、"包含数据图表"等）
+"""
+
+            messages = [
+                {"role": "system", "content": "你是一个专业的数据提取助手。请严格按JSON格式输出。"},
+                {"role": "user", "content": prompt}
+            ]
+
+            response = await self.llm.chat(
+                messages=messages,
+                temperature=0.1,
+                max_tokens=50000
+            )
+
+            content = self.llm.extract_message_content(response)
+
+            # 解析 JSON
+            result = self._parse_json_response(content)
+
+            if result:
+                logger.info(f"AI 表格提取成功: {len(result.get('rows', []))} 行数据")
+                return {
+                    "success": True,
+                    "type": "table_data",
+                    "headers": result.get("headers", []),
+                    "rows": result.get("rows", []),
+                    "description": result.get("description", "")
+                }
+            else:
+                # 如果 AI 返回格式不对，尝试直接解析表格
+                return self._fallback_table_parse(tables)
+
+        except Exception as e:
+            logger.error(f"AI 表格提取失败: {str(e)}")
+            return self._fallback_table_parse(tables)
+
+    async def _extract_from_text_with_ai(
+        self,
+        paragraphs: List[str],
+        full_text: str,
+        image_count: int,
+        image_descriptions: List[str],
+        user_hint: str,
+        image_analysis: str = ""
+    ) -> Dict[str, Any]:
+        """
+        使用 AI 从 Word 纯文本中提取结构化数据
+
+        Args:
+            paragraphs: 段落列表
+            full_text: 完整文本
+            image_count: 图片数量
+            image_descriptions: 图片描述列表
+            user_hint: 用户提示
+            image_analysis: 图片 AI 分析结果
+
+        Returns:
+            结构化数据
+        """
+        try:
+            # 限制文本长度
+            text_preview = full_text[:8000] if len(full_text) > 8000 else full_text
+
+            # 图片提示
+            image_hint = f"\n【文档图片】此文档包含 {image_count} 张图片/图表。" if image_count > 0 else ""
+            if image_descriptions:
+                image_hint += "\n" + "\n".join(image_descriptions)
+
+            prompt = f"""你是一个专业的数据提取专家。请从以下 Word 文档的完整内容中提取结构化数据。
+
+【用户需求】
+{user_hint if user_hint else "请识别并提取文档中的关键信息，包括：表格数据、键值对、列表项等。"}
+
+【文档正文】{image_hint}
+{text_preview}
+
+请按照以下 JSON 格式输出：
+{{
+    "type": "structured_text",
+    "tables": [{{"headers": [...], "rows": [...]}}],
+    "key_values": {{"键1": "值1", "键2": "值2", ...}},
+    "list_items": ["项1", "项2", ...],
+    "summary": "文档内容摘要"
+}}
+
+重点：
+- 如果文档包含表格数据，提取到 tables 中
+- 如果文档包含键值对（如 名称: 张三），提取到 key_values 中
+- 如果文档包含列表项，提取到 list_items 中
+- 如果文档包含图片，请根据上下文推断图片内容（如"流程图"、"数据折线图"等）并在 description 中说明
+- 如果无法提取到结构化数据，至少提供一个详细的摘要
+"""
+
+            messages = [
+                {"role": "system", "content": "你是一个专业的数据提取助手。请严格按JSON格式输出。"},
+                {"role": "user", "content": prompt}
+            ]
+
+            response = await self.llm.chat(
+                messages=messages,
+                temperature=0.1,
+                max_tokens=50000
+            )
+
+            content = self.llm.extract_message_content(response)
+
+            result = self._parse_json_response(content)
+
+            if result:
+                logger.info(f"AI 文本提取成功: type={result.get('type')}")
+                return {
+                    "success": True,
+                    "type": result.get("type", "structured_text"),
+                    "tables": result.get("tables", []),
+                    "key_values": result.get("key_values", {}),
+                    "list_items": result.get("list_items", []),
+                    "summary": result.get("summary", ""),
+                    "raw_text_preview": text_preview[:500]
+                }
+            else:
+                return {
+                    "success": True,
+                    "type": "text",
+                    "summary": text_preview[:500],
+                    "raw_text_preview": text_preview[:500]
+                }
+
+        except Exception as e:
+            logger.error(f"AI 文本提取失败: {str(e)}")
+            return {
+                "success": False,
+                "error": str(e)
+            }
+
+    async def _analyze_images_with_ai(
+        self,
+        images: List[Dict[str, str]],
+        user_hint: str = ""
+    ) -> str:
+        """
+        使用视觉模型分析 Word 文档中的图片
+
+        Args:
+            images: 图片列表，每项包含 base64 和 mime_type
+            user_hint: 用户提示
+
+        Returns:
+            图片分析结果文本
+        """
+        try:
+            # 调用 LLM 的视觉分析功能
+            result = await self.llm.analyze_images(
+                images=images,
+                user_prompt=user_hint or "请详细描述图片内容，提取所有文字和数据信息。"
+            )
+
+            if result.get("success"):
+                analysis = result.get("analysis", {})
+                if isinstance(analysis, dict):
+                    description = analysis.get("description", "")
+                    text_content = analysis.get("text_content", "")
+                    data_extracted = analysis.get("data_extracted", {})
+
+                    result_text = f"【图片分析结果】\n{description}"
+                    if text_content:
+                        result_text += f"\n\n【图片中的文字】\n{text_content}"
+                    if data_extracted:
+                        result_text += f"\n\n【提取的数据】\n{json.dumps(data_extracted, ensure_ascii=False)}"
+                    return result_text
+                else:
+                    return str(analysis)
+            else:
+                logger.warning(f"图片 AI 分析失败: {result.get('error')}")
+                return ""
+
+        except Exception as e:
+            logger.error(f"图片 AI 分析异常: {str(e)}")
+            return ""
+
+    def _build_tables_description(self, tables: List[Dict]) -> str:
+        """构建表格的文本描述"""
+        result = []
+
+        for idx, table in enumerate(tables):
+            rows = table.get("rows", [])
+            if not rows:
+                continue
+
+            result.append(f"\n--- 表格 {idx + 1} ---")
+
+            for row_idx, row in enumerate(rows[:50]):  # 限制每表格最多50行
+                if isinstance(row, list):
+                    result.append(" | ".join(str(cell).strip() for cell in row))
+                elif isinstance(row, dict):
+                    result.append(str(row))
+
+            if len(rows) > 50:
+                result.append(f"...（共 {len(rows)} 行，仅显示前50行）")
+
+        return "\n".join(result) if result else "（无表格内容）"
+
+    def _parse_json_response(self, content: str) -> Optional[Dict]:
+        """解析 JSON 响应，处理各种格式问题"""
+        import re
+
+        # 清理 markdown 标记
+        cleaned = content.strip()
+        cleaned = re.sub(r'^```json\s*', '', cleaned, flags=re.MULTILINE)
+        cleaned = re.sub(r'^```\s*', '', cleaned, flags=re.MULTILINE)
+        cleaned = cleaned.strip()
+
+        # 找到 JSON 开始位置
+        json_start = -1
+        for i, c in enumerate(cleaned):
+            if c == '{':
+                json_start = i
+                break
+
+        if json_start == -1:
+            logger.warning("无法找到 JSON 开始位置")
+            return None
+
+        json_text = cleaned[json_start:]
+
+        # 尝试直接解析
+        try:
+            return json.loads(json_text)
+        except json.JSONDecodeError:
+            pass
+
+        # 尝试修复并解析
+        try:
+            # 找到闭合括号
+            depth = 0
+            end_pos = -1
+            for i, c in enumerate(json_text):
+                if c == '{':
+                    depth += 1
+                elif c == '}':
+                    depth -= 1
+                    if depth == 0:
+                        end_pos = i + 1
+                        break
+
+            if end_pos > 0:
+                fixed = json_text[:end_pos]
+                # 移除末尾逗号
+                fixed = re.sub(r',\s*([}]])', r'\1', fixed)
+                return json.loads(fixed)
+        except Exception as e:
+            logger.warning(f"JSON 修复失败: {e}")
+
+        return None
+
+    def _fallback_table_parse(self, tables: List[Dict]) -> Dict[str, Any]:
+        """当 AI 解析失败时，直接解析表格"""
+        if not tables:
+            return {
+                "success": True,
+                "type": "empty",
+                "data": {},
+                "message": "无表格内容"
+            }
+
+        all_rows = []
+        all_headers = None
+
+        for table in tables:
+            rows = table.get("rows", [])
+            if not rows:
+                continue
+
+            # 查找真正的表头行（跳过标题行）
+            header_row_idx = 0
+            for idx, row in enumerate(rows[:5]):  # 只检查前5行
+                if not isinstance(row, list):
+                    continue
+                # 如果某一行包含"表"字开头且单元格内容很长，这可能是标题行
+                first_cell = str(row[0]) if row else ""
+                if first_cell.startswith("表") and len(first_cell) > 15:
+                    header_row_idx = idx + 1
+                    continue
+                # 如果某一行有超过3个空单元格，可能是无效行
+                empty_count = sum(1 for cell in row if not str(cell).strip())
+                if empty_count > 3:
+                    header_row_idx = idx + 1
+                    continue
+                # 找到第一行看起来像表头的行（短单元格，大部分有内容）
+                avg_len = sum(len(str(c)) for c in row) / len(row) if row else 0
+                if avg_len < 20:  # 表头通常比数据行短
+                    header_row_idx = idx
+                    break
+
+            if header_row_idx >= len(rows):
+                continue
+
+            # 使用找到的表头行
+            if rows and isinstance(rows[header_row_idx], list):
+                headers = rows[header_row_idx]
+                if all_headers is None:
+                    all_headers = headers
+
+                # 数据行（从表头之后开始）
+                for row in rows[header_row_idx + 1:]:
+                    if isinstance(row, list) and len(row) == len(headers):
+                        all_rows.append(row)
+
+        if all_headers and all_rows:
+            return {
+                "success": True,
+                "type": "table_data",
+                "headers": all_headers,
+                "rows": all_rows,
+                "description": "直接从 Word 表格提取"
+            }
+
+        return {
+            "success": True,
+            "type": "raw",
+            "tables": tables,
+            "message": "表格数据（未AI处理）"
+        }
+
+    async def fill_template_with_ai(
+        self,
+        file_path: str,
+        template_fields: List[Dict[str, Any]],
+        user_hint: str = ""
+    ) -> Dict[str, Any]:
+        """
+        使用 AI 解析 Word 文档并填写模板
+
+        这是主要入口函数，前端调用此函数即可完成：
+        1. AI 解析 Word 文档
+        2. 根据模板字段提取数据
+        3. 返回填写结果
+
+        Args:
+            file_path: Word 文件路径
+            template_fields: 模板字段列表 [{"name": "字段名", "hint": "提示词"}, ...]
+            user_hint: 用户提示
+
+        Returns:
+            填写结果
+        """
+        try:
+            # 1. AI 解析文档
+            parse_result = await self.parse_word_with_ai(file_path, user_hint)
+
+            if not parse_result.get("success"):
+                return {
+                    "success": False,
+                    "error": parse_result.get("error", "解析失败"),
+                    "filled_data": {},
+                    "source": "ai_parse_failed"
+                }
+
+            # 2. 根据字段类型提取数据
+            filled_data = {}
+            extract_details = []
+
+            parse_type = parse_result.get("type", "")
+
+            if parse_type == "table_data":
+                # 表格数据：直接匹配列名
+                headers = parse_result.get("headers", [])
+                rows = parse_result.get("rows", [])
+
+                for field in template_fields:
+                    field_name = field.get("name", "")
+                    values = self._extract_field_from_table(headers, rows, field_name)
+                    filled_data[field_name] = values
+                    extract_details.append({
+                        "field": field_name,
+                        "values": values,
+                        "source": "ai_table_extraction",
+                        "confidence": 0.9 if values else 0.0
+                    })
+
+            elif parse_type == "structured_text":
+                # 结构化文本：尝试从 key_values 和 list_items 提取
+                key_values = parse_result.get("key_values", {})
+                list_items = parse_result.get("list_items", [])
+
+                for field in template_fields:
+                    field_name = field.get("name", "")
+                    value = key_values.get(field_name, "")
+                    if not value and list_items:
+                        value = list_items[0] if list_items else ""
+                    filled_data[field_name] = [value] if value else []
+                    extract_details.append({
+                        "field": field_name,
+                        "values": [value] if value else [],
+                        "source": "ai_text_extraction",
+                        "confidence": 0.7 if value else 0.0
+                    })
+
+            else:
+                # 其他类型：返回原始解析结果供后续处理
+                for field in template_fields:
+                    field_name = field.get("name", "")
+                    filled_data[field_name] = []
+                    extract_details.append({
+                        "field": field_name,
+                        "values": [],
+                        "source": "no_ai_data",
+                        "confidence": 0.0
+                    })
+
+            # 3. 返回结果
+            max_rows = max(len(v) for v in filled_data.values()) if filled_data else 1
+
+            return {
+                "success": True,
+                "filled_data": filled_data,
+                "fill_details": extract_details,
+                "ai_parse_result": {
+                    "type": parse_type,
+                    "description": parse_result.get("description", "")
+                },
+                "source_doc_count": 1,
+                "max_rows": max_rows
+            }
+
+        except Exception as e:
+            logger.error(f"AI 填表失败: {str(e)}")
+            return {
+                "success": False,
+                "error": str(e),
+                "filled_data": {},
+                "fill_details": []
+            }
+
+    def _extract_field_from_table(
+        self,
+        headers: List[str],
+        rows: List[List],
+        field_name: str
+    ) -> List[str]:
+        """从表格中提取指定字段的值"""
+        # 查找匹配的列
+        target_col_idx = None
+        for col_idx, header in enumerate(headers):
+            if field_name.lower() in str(header).lower() or str(header).lower() in field_name.lower():
+                target_col_idx = col_idx
+                break
+
+        if target_col_idx is None:
+            return []
+
+        # 提取该列所有值
+        values = []
+        for row in rows:
+            if isinstance(row, list) and target_col_idx < len(row):
+                val = str(row[target_col_idx]).strip()
+                if val:
+                    values.append(val)
+
+        return values
+
+
+# 全局单例
+word_ai_service = WordAIService()

backend/requirements.txt

@@ -1,4 +1,4 @@
-# ============================================================
+# ============================================================
 # 基于大语言模型的文档理解与多源数据融合系统
 # Python 依赖清单
 # ============================================================

frontend/src/App.tsx

@@ -1,5 +1,5 @@
 import { RouterProvider } from 'react-router-dom';
-import { AuthProvider } from '@/context/AuthContext';
+import { AuthProvider } from '@/contexts/AuthContext';
 import { TemplateFillProvider } from '@/context/TemplateFillContext';
 import { router } from '@/routes';
 import { Toaster } from 'sonner';

frontend/src/components/common/RouteGuard.tsx

@@ -1,6 +1,6 @@
 import React from 'react';
 import { Navigate, useLocation } from 'react-router-dom';
-import { useAuth } from '@/context/AuthContext';
+import { useAuth } from '@/contexts/AuthContext';
 
 export const RouteGuard: React.FC<{ children: React.ReactNode }> = ({ children }) => {
   const { user, loading } = useAuth();

frontend/src/context/AuthContext.tsx

@@ -1,85 +0,0 @@
-import React, { createContext, useContext, useEffect, useState } from 'react';
-import { supabase } from '@/db/supabase';
-import { User } from '@supabase/supabase-js';
-import { Profile } from '@/types/types';
-
-interface AuthContextType {
-  user: User | null;
-  profile: Profile | null;
-  signIn: (email: string, password: string) => Promise<{ error: any }>;
-  signUp: (email: string, password: string) => Promise<{ error: any }>;
-  signOut: () => Promise<{ error: any }>;
-  loading: boolean;
-}
-
-const AuthContext = createContext<AuthContextType | undefined>(undefined);
-
-export const AuthProvider: React.FC<{ children: React.ReactNode }> = ({ children }) => {
-  const [user, setUser] = useState<User | null>(null);
-  const [profile, setProfile] = useState<Profile | null>(null);
-  const [loading, setLoading] = useState(true);
-
-  useEffect(() => {
-    // Check active sessions and sets the user
-    supabase.auth.getSession().then(({ data: { session } }) => {
-      setUser(session?.user ?? null);
-      if (session?.user) fetchProfile(session.user.id);
-      else setLoading(false);
-    });
-
-    // Listen for changes on auth state (sign in, sign out, etc.)
-    const { data: { subscription } } = supabase.auth.onAuthStateChange((_event, session) => {
-      setUser(session?.user ?? null);
-      if (session?.user) fetchProfile(session.user.id);
-      else {
-        setProfile(null);
-        setLoading(false);
-      }
-    });
-
-    return () => subscription.unsubscribe();
-  }, []);
-
-  const fetchProfile = async (uid: string) => {
-    try {
-      const { data, error } = await supabase
-        .from('profiles')
-        .select('*')
-        .eq('id', uid)
-        .maybeSingle();
-      
-      if (error) throw error;
-      setProfile(data);
-    } catch (err) {
-      console.error('Error fetching profile:', err);
-    } finally {
-      setLoading(false);
-    }
-  };
-
-  const signIn = async (email: string, password: string) => {
-    return await supabase.auth.signInWithPassword({ email, password });
-  };
-
-  const signUp = async (email: string, password: string) => {
-    return await supabase.auth.signUp({ email, password });
-  };
-
-  const signOut = async () => {
-    return await supabase.auth.signOut();
-  };
-
-  return (
-    <AuthContext.Provider value={{ user, profile, signIn, signUp, signOut, loading }}>
-      {children}
-    </AuthContext.Provider>
-  );
-};
-
-export const useAuth = () => {
-  const context = useContext(AuthContext);
-  if (context === undefined) {
-    throw new Error('useAuth must be used within an AuthProvider');
-  }
-  return context;
-};

frontend/src/db/backend-api.ts

@@ -400,6 +400,49 @@ export const backendApi = {
     }
   },
 
+  /**
+   * 获取任务历史列表
+   */
+  async getTasks(
+    limit: number = 50,
+    skip: number = 0
+  ): Promise<{ success: boolean; tasks: any[]; count: number }> {
+    const url = `${BACKEND_BASE_URL}/tasks?limit=${limit}&skip=${skip}`;
+
+    try {
+      const response = await fetch(url);
+      if (!response.ok) {
+        const error = await response.json();
+        throw new Error(error.detail || '获取任务列表失败');
+      }
+      return await response.json();
+    } catch (error) {
+      console.error('获取任务列表失败:', error);
+      throw error;
+    }
+  },
+
+  /**
+   * 删除任务
+   */
+  async deleteTask(taskId: string): Promise<{ success: boolean; deleted: boolean }> {
+    const url = `${BACKEND_BASE_URL}/tasks/${taskId}`;
+
+    try {
+      const response = await fetch(url, {
+        method: 'DELETE'
+      });
+      if (!response.ok) {
+        const error = await response.json();
+        throw new Error(error.detail || '删除任务失败');
+      }
+      return await response.json();
+    } catch (error) {
+      console.error('删除任务失败:', error);
+      throw error;
+    }
+  },
+
   /**
    * 轮询任务状态直到完成
    */
@@ -764,6 +807,41 @@ export const backendApi = {
     }
   },
 
+  /**
+   * 填充原始模板并导出
+   *
+   * 直接打开原始模板文件，将数据填入模板的表格/单元格中，然后导出
+   * 适用于比赛场景：保持原始模板格式不变
+   */
+  async fillAndExportTemplate(
+    templatePath: string,
+    filledData: Record<string, any>,
+    format: 'xlsx' | 'docx' = 'xlsx'
+  ): Promise<Blob> {
+    const url = `${BACKEND_BASE_URL}/templates/fill-and-export`;
+
+    try {
+      const response = await fetch(url, {
+        method: 'POST',
+        headers: { 'Content-Type': 'application/json' },
+        body: JSON.stringify({
+          template_path: templatePath,
+          filled_data: filledData,
+          format,
+        }),
+      });
+
+      if (!response.ok) {
+        const error = await response.json();
+        throw new Error(error.detail || '填充模板失败');
+      }
+      return await response.blob();
+    } catch (error) {
+      console.error('填充模板失败:', error);
+      throw error;
+    }
+  },
+
   // ==================== Excel 专用接口 (保留兼容) ====================
 
   /**
@@ -1145,7 +1223,7 @@ export const aiApi = {
 
     try {
       const response = await fetch(url, {
-        method: 'GET',
+        method: 'POST',
         body: formData,
       });
 
@@ -1161,6 +1239,48 @@ export const aiApi = {
     }
   },
 
+  /**
+   * 上传并使用 AI 分析 TXT 文本文件，提取结构化数据
+   */
+  async analyzeTxt(
+    file: File
+  ): Promise<{
+    success: boolean;
+    filename?: string;
+    structured_data?: {
+      table?: {
+        columns?: string[];
+        rows?: string[][];
+      };
+      summary?: string;
+      key_value_pairs?: Array<{ key: string; value: string }>;
+      numeric_data?: Array<{ name: string; value: number; unit?: string }>;
+    };
+    error?: string;
+  }> {
+    const formData = new FormData();
+    formData.append('file', file);
+
+    const url = `${BACKEND_BASE_URL}/ai/analyze/txt`;
+
+    try {
+      const response = await fetch(url, {
+        method: 'POST',
+        body: formData,
+      });
+
+      if (!response.ok) {
+        const error = await response.json();
+        throw new Error(error.detail || 'TXT AI 分析失败');
+      }
+
+      return await response.json();
+    } catch (error) {
+      console.error('TXT AI 分析失败:', error);
+      throw error;
+    }
+  },
+
   /**
    * 生成统计信息和图表
    */
@@ -1259,4 +1379,84 @@ export const aiApi = {
       throw error;
     }
   },
+
+  // ==================== Word AI 解析 ====================
+
+  /**
+   * 使用 AI 解析 Word 文档，提取结构化数据
+   */
+  async analyzeWordWithAI(
+    file: File,
+    userHint: string = ''
+  ): Promise<{
+    success: boolean;
+    type?: string;
+    headers?: string[];
+    rows?: string[][];
+    key_values?: Record<string, string>;
+    list_items?: string[];
+    summary?: string;
+    error?: string;
+  }> {
+    const formData = new FormData();
+    formData.append('file', file);
+    if (userHint) {
+      formData.append('user_hint', userHint);
+    }
+
+    const url = `${BACKEND_BASE_URL}/ai/analyze/word`;
+
+    try {
+      const response = await fetch(url, {
+        method: 'POST',
+        body: formData,
+      });
+
+      if (!response.ok) {
+        const error = await response.json();
+        throw new Error(error.detail || 'Word AI 解析失败');
+      }
+
+      return await response.json();
+    } catch (error) {
+      console.error('Word AI 解析失败:', error);
+      throw error;
+    }
+  },
+
+  /**
+   * 使用 AI 解析 Word 文档并填写模板
+   * 一次性完成：AI解析 + 填表
+   */
+  async fillTemplateFromWordAI(
+    file: File,
+    templateFields: TemplateField[],
+    userHint: string = ''
+  ): Promise<FillResult> {
+    const formData = new FormData();
+    formData.append('file', file);
+    formData.append('template_fields', JSON.stringify(templateFields));
+    if (userHint) {
+      formData.append('user_hint', userHint);
+    }
+
+    const url = `${BACKEND_BASE_URL}/ai/analyze/word/fill-template`;
+
+    try {
+      const response = await fetch(url, {
+        method: 'POST',
+        body: formData,
+      });
+
+      if (!response.ok) {
+        const error = await response.json();
+        throw new Error(error.detail || 'Word AI 填表失败');
+      }
+
+      return await response.json();
+    } catch (error) {
+      console.error('Word AI 填表失败:', error);
+      throw error;
+    }
+  },
 };

frontend/src/pages/Documents.tsx

@@ -766,6 +766,7 @@ const Documents: React.FC = () => {
                         <div
                           {...getRootProps()}
                           className="flex items-center justify-center gap-2 p-3 border-2 border-dashed rounded-lg cursor-pointer hover:border-primary/50 hover:bg-primary/5 transition-colors"
+                          onClick={(e) => e.stopPropagation()}
                         >
                           <input {...getInputProps()} multiple={true} />
                           <Plus size={16} className="text-muted-foreground" />

frontend/src/pages/FormFill.tsx

@@ -1,603 +0,0 @@
-import React, { useState, useEffect } from 'react';
-import {
-  TableProperties,
-  Plus,
-  FilePlus,
-  CheckCircle2,
-  Download,
-  Clock,
-  RefreshCcw,
-  Sparkles,
-  Zap,
-  FileCheck,
-  FileSpreadsheet,
-  Trash2,
-  ChevronDown,
-  ChevronUp,
-  BarChart3,
-  FileText,
-  TrendingUp,
-  Info,
-  AlertCircle,
-  Loader2
-} from 'lucide-react';
-import { Button } from '@/components/ui/button';
-import { Card, CardContent, CardHeader, CardTitle, CardDescription, CardFooter } from '@/components/ui/card';
-import { Badge } from '@/components/ui/badge';
-import { useAuth } from '@/context/AuthContext';
-import { templateApi, documentApi, taskApi } from '@/db/api';
-import { backendApi, aiApi } from '@/db/backend-api';
-import { supabase } from '@/db/supabase';
-import { format } from 'date-fns';
-import { toast } from 'sonner';
-import { cn } from '@/lib/utils';
-import { Skeleton } from '@/components/ui/skeleton';
-import {
-  Dialog,
-  DialogContent,
-  DialogHeader,
-  DialogTitle,
-  DialogTrigger,
-  DialogFooter,
-  DialogDescription
-} from '@/components/ui/dialog';
-import { Checkbox } from '@/components/ui/checkbox';
-import { ScrollArea } from '@/components/ui/scroll-area';
-import { Input } from '@/components/ui/input';
-import { Label } from '@/components/ui/label';
-import { Textarea } from '@/components/ui/textarea';
-import { Select, SelectContent, SelectItem, SelectTrigger, SelectValue } from '@/components/ui/select';
-import { useDropzone } from 'react-dropzone';
-import { Markdown } from '@/components/ui/markdown';
-
-type Template = any;
-type Document = any;
-type FillTask = any;
-
-const FormFill: React.FC = () => {
-  const { profile } = useAuth();
-  const [templates, setTemplates] = useState<Template[]>([]);
-  const [documents, setDocuments] = useState<Document[]>([]);
-  const [tasks, setTasks] = useState<any[]>([]);
-  const [loading, setLoading] = useState(true);
-
-  // Selection state
-  const [selectedTemplate, setSelectedTemplate] = useState<string | null>(null);
-  const [selectedDocs, setSelectedDocs] = useState<string[]>([]);
-  const [creating, setCreating] = useState(false);
-  const [openTaskDialog, setOpenTaskDialog] = useState(false);
-  const [viewingTask, setViewingTask] = useState<any | null>(null);
-
-  // Excel upload state
-  const [excelFile, setExcelFile] = useState<File | null>(null);
-  const [excelParseResult, setExcelParseResult] = useState<any>(null);
-  const [excelAnalysis, setExcelAnalysis] = useState<any>(null);
-  const [excelAnalyzing, setExcelAnalyzing] = useState(false);
-  const [expandedSheet, setExpandedSheet] = useState<string | null>(null);
-  const [aiOptions, setAiOptions] = useState({
-    userPrompt: '请分析这些数据，并提取关键信息用于填表，包括数值、分类、摘要等。',
-    analysisType: 'general' as 'general' | 'summary' | 'statistics' | 'insights'
-  });
-
-  const loadData = async () => {
-    if (!profile) return;
-    try {
-      const [t, d, ts] = await Promise.all([
-        templateApi.listTemplates((profile as any).id),
-        documentApi.listDocuments((profile as any).id),
-        taskApi.listTasks((profile as any).id)
-      ]);
-      setTemplates(t);
-      setDocuments(d);
-      setTasks(ts);
-    } catch (err: any) {
-      toast.error('数据加载失败');
-    } finally {
-      setLoading(false);
-    }
-  };
-
-  useEffect(() => {
-    loadData();
-  }, [profile]);
-
-  // Excel upload handlers
-  const onExcelDrop = async (acceptedFiles: File[]) => {
-    const file = acceptedFiles[0];
-    if (!file) return;
-
-    if (!file.name.match(/\.(xlsx|xls)$/i)) {
-      toast.error('仅支持 .xlsx 和 .xls 格式的 Excel 文件');
-      return;
-    }
-
-    setExcelFile(file);
-    setExcelParseResult(null);
-    setExcelAnalysis(null);
-    setExpandedSheet(null);
-
-    try {
-      const result = await backendApi.uploadExcel(file);
-      if (result.success) {
-        toast.success(`Excel 解析成功: ${file.name}`);
-        setExcelParseResult(result);
-      } else {
-        toast.error(result.error || '解析失败');
-      }
-    } catch (error: any) {
-      toast.error(error.message || '上传失败');
-    }
-  };
-
-  const { getRootProps, getInputProps, isDragActive } = useDropzone({
-    onDrop: onExcelDrop,
-    accept: {
-      'application/vnd.openxmlformats-officedocument.spreadsheetml.sheet': ['.xlsx'],
-      'application/vnd.ms-excel': ['.xls']
-    },
-    maxFiles: 1
-  });
-
-  const handleAnalyzeExcel = async () => {
-    if (!excelFile || !excelParseResult?.success) {
-      toast.error('请先上传并解析 Excel 文件');
-      return;
-    }
-
-    setExcelAnalyzing(true);
-    setExcelAnalysis(null);
-
-    try {
-      const result = await aiApi.analyzeExcel(excelFile, {
-        userPrompt: aiOptions.userPrompt,
-        analysisType: aiOptions.analysisType
-      });
-
-      if (result.success) {
-        toast.success('AI 分析完成');
-        setExcelAnalysis(result);
-      } else {
-        toast.error(result.error || 'AI 分析失败');
-      }
-    } catch (error: any) {
-      toast.error(error.message || 'AI 分析失败');
-    } finally {
-      setExcelAnalyzing(false);
-    }
-  };
-
-  const handleUseExcelData = () => {
-    if (!excelParseResult?.success) {
-      toast.error('请先解析 Excel 文件');
-      return;
-    }
-
-    // 将 Excel 解析的数据标记为"文档"，添加到选择列表
-    toast.success('Excel 数据已添加到数据源，请在任务对话框中选择');
-    // 这里可以添加逻辑来将 Excel 数据传递给后端创建任务
-  };
-
-  const handleDeleteExcel = () => {
-    setExcelFile(null);
-    setExcelParseResult(null);
-    setExcelAnalysis(null);
-    setExpandedSheet(null);
-    toast.success('Excel 文件已清除');
-  };
-
-  const handleUploadTemplate = async (e: React.ChangeEvent<HTMLInputElement>) => {
-    const file = e.target.files?.[0];
-    if (!file || !profile) return;
-
-    try {
-      toast.loading('正在上传模板...');
-      await templateApi.uploadTemplate(file, (profile as any).id);
-      toast.dismiss();
-      toast.success('模板上传成功');
-      loadData();
-    } catch (err) {
-      toast.dismiss();
-      toast.error('上传模板失败');
-    }
-  };
-
-  const handleCreateTask = async () => {
-    if (!profile || !selectedTemplate || selectedDocs.length === 0) {
-      toast.error('请先选择模板和数据源文档');
-      return;
-    }
-
-    setCreating(true);
-    try {
-      const task = await taskApi.createTask((profile as any).id, selectedTemplate, selectedDocs);
-      if (task) {
-        toast.success('任务已创建，正在进行智能填表...');
-        setOpenTaskDialog(false);
-
-        // Invoke edge function
-        supabase.functions.invoke('fill-template', {
-          body: { taskId: task.id }
-        }).then(({ error }) => {
-          if (error) toast.error('填表任务执行失败');
-          else {
-            toast.success('表格填写完成！');
-            loadData();
-          }
-        });
-        loadData();
-      }
-    } catch (err: any) {
-      toast.error('创建任务失败');
-    } finally {
-      setCreating(false);
-    }
-  };
-
-  const getStatusColor = (status: string) => {
-    switch (status) {
-      case 'completed': return 'bg-emerald-500 text-white';
-      case 'failed': return 'bg-destructive text-white';
-      default: return 'bg-amber-500 text-white';
-    }
-  };
-
-  const formatFileSize = (bytes: number): string => {
-    if (bytes === 0) return '0 B';
-    const k = 1024;
-    const sizes = ['B', 'KB', 'MB', 'GB'];
-    const i = Math.floor(Math.log(bytes) / Math.log(k));
-    return `${(bytes / Math.pow(k, i)).toFixed(2)} ${sizes[i]}`;
-  };
-
-  return (
-    <div className="space-y-8 animate-fade-in pb-10">
-      <section className="flex flex-col md:flex-row md:items-center justify-between gap-4">
-        <div className="space-y-1">
-          <h1 className="text-3xl font-extrabold tracking-tight">智能填表</h1>
-          <p className="text-muted-foreground">根据您的表格模板，自动聚合多源文档信息进行精准填充，告别重复劳动。</p>
-        </div>
-        <div className="flex items-center gap-3">
-          <Dialog open={openTaskDialog} onOpenChange={setOpenTaskDialog}>
-            <DialogTrigger asChild>
-              <Button className="rounded-xl shadow-lg shadow-primary/20 gap-2 h-11 px-6">
-                <FilePlus size={18} />
-                <span>新建填表任务</span>
-              </Button>
-            </DialogTrigger>
-            <DialogContent className="max-w-4xl max-h-[90vh] flex flex-col p-0 overflow-hidden border-none shadow-2xl rounded-3xl">
-              <DialogHeader className="p-8 pb-4 bg-muted/50">
-                <DialogTitle className="text-2xl font-bold flex items-center gap-2">
-                  <Sparkles size={24} className="text-primary" />
-                  开启智能填表之旅
-                </DialogTitle>
-                <DialogDescription>
-                  选择一个表格模板及若干个数据源文档，AI 将自动为您分析并填写。
-                </DialogDescription>
-              </DialogHeader>
-
-              <ScrollArea className="flex-1 p-8 pt-4">
-                <div className="space-y-8">
-                  {/* Step 1: Select Template */}
-                  <div className="space-y-4">
-                    <div className="flex items-center justify-between">
-                      <h4 className="font-bold flex items-center gap-2 text-primary uppercase tracking-widest text-xs">
-                        <span className="w-5 h-5 rounded-full bg-primary text-white flex items-center justify-center text-[10px]">1</span>
-                        选择表格模板
-                      </h4>
-                      <label className="cursor-pointer text-xs font-semibold text-primary hover:underline flex items-center gap-1">
-                        <Plus size={12} /> 上传新模板
-                        <input type="file" className="hidden" onChange={handleUploadTemplate} accept=".docx,.xlsx" />
-                      </label>
-                    </div>
-                    {templates.length > 0 ? (
-                      <div className="grid grid-cols-1 sm:grid-cols-2 gap-3">
-                        {templates.map(t => (
-                          <div
-                            key={t.id}
-                            className={cn(
-                              "p-4 rounded-2xl border-2 transition-all cursor-pointer flex items-center gap-3 group relative overflow-hidden",
-                              selectedTemplate === t.id ? "border-primary bg-primary/5" : "border-border hover:border-primary/50"
-                            )}
-                            onClick={() => setSelectedTemplate(t.id)}
-                          >
-                            <div className={cn(
-                              "w-10 h-10 rounded-xl flex items-center justify-center shrink-0 transition-colors",
-                              selectedTemplate === t.id ? "bg-primary text-white" : "bg-muted text-muted-foreground"
-                            )}>
-                              <TableProperties size={20} />
-                            </div>
-                            <div className="flex-1 min-w-0">
-                              <p className="font-bold text-sm truncate">{t.name}</p>
-                              <p className="text-[10px] text-muted-foreground uppercase">{t.type}</p>
-                            </div>
-                            {selectedTemplate === t.id && (
-                              <div className="absolute top-0 right-0 w-8 h-8 bg-primary text-white flex items-center justify-center rounded-bl-xl">
-                                <CheckCircle2 size={14} />
-                              </div>
-                            )}
-                          </div>
-                        ))}
-                      </div>
-                    ) : (
-                      <div className="p-8 text-center bg-muted/30 rounded-2xl border border-dashed text-sm italic text-muted-foreground">
-                        暂无模板，请先点击右上角上传。
-                      </div>
-                    )}
-                  </div>
-
-                  {/* Step 2: Upload & Analyze Excel */}
-                  <div className="space-y-4">
-                    <h4 className="font-bold flex items-center gap-2 text-primary uppercase tracking-widest text-xs">
-                      <span className="w-5 h-5 rounded-full bg-primary text-white flex items-center justify-center text-[10px]">1.5</span>
-                      Excel 数据源
-                    </h4>
-                    <div className="bg-muted/20 rounded-2xl p-6">
-                      {!excelFile ? (
-                        <div
-                          {...getRootProps()}
-                          className={cn(
-                            "border-2 border-dashed rounded-xl p-8 transition-all duration-300 flex flex-col items-center justify-center text-center cursor-pointer group",
-                            isDragActive ? "border-primary bg-primary/5" : "border-muted-foreground/20 hover:border-primary/50 hover:bg-muted/30"
-                          )}
-                        >
-                          <input {...getInputProps()} />
-                          <div className="w-12 h-12 rounded-xl bg-primary/10 text-primary flex items-center justify-center mb-3 group-hover:scale-110 transition-transform">
-                            <FileSpreadsheet size={24} />
-                          </div>
-                          <p className="font-semibold text-sm">
-                            {isDragActive ? '释放以开始上传' : '点击或拖拽 Excel 文件'}
-                          </p>
-                          <p className="text-xs text-muted-foreground mt-1">支持 .xlsx 和 .xls 格式</p>
-                        </div>
-                      ) : (
-                        <div className="space-y-4">
-                          <div className="flex items-center gap-3 p-3 bg-background rounded-xl">
-                            <div className="w-10 h-10 rounded-lg bg-emerald-500/10 text-emerald-500 flex items-center justify-center">
-                              <FileSpreadsheet size={20} />
-                            </div>
-                            <div className="flex-1 min-w-0">
-                              <p className="font-semibold text-sm truncate">{excelFile.name}</p>
-                              <p className="text-xs text-muted-foreground">{formatFileSize(excelFile.size)}</p>
-                            </div>
-                            <div className="flex gap-2">
-                              <Button
-                                variant="ghost"
-                                size="icon"
-                                className="text-destructive hover:bg-destructive/10"
-                                onClick={handleDeleteExcel}
-                              >
-                                <Trash2 size={16} />
-                              </Button>
-                            </div>
-                          </div>
-
-                          {/* AI Analysis Options */}
-                          {excelParseResult?.success && (
-                            <div className="space-y-3">
-                              <div className="space-y-2">
-                                <Label htmlFor="analysis-type" className="text-xs">分析类型</Label>
-                                <Select
-                                  value={aiOptions.analysisType}
-                                  onValueChange={(value: any) => setAiOptions({ ...aiOptions, analysisType: value })}
-                                >
-                                  <SelectTrigger id="analysis-type" className="bg-background h-9 text-sm">
-                                    <SelectValue placeholder="选择分析类型" />
-                                  </SelectTrigger>
-                                  <SelectContent>
-                                    <SelectItem value="general">综合分析</SelectItem>
-                                    <SelectItem value="summary">数据摘要</SelectItem>
-                                    <SelectItem value="statistics">统计分析</SelectItem>
-                                    <SelectItem value="insights">深度洞察</SelectItem>
-                                  </SelectContent>
-                                </Select>
-                              </div>
-                              <div className="space-y-2">
-                                <Label htmlFor="user-prompt" className="text-xs">自定义提示词</Label>
-                                <Textarea
-                                  id="user-prompt"
-                                  value={aiOptions.userPrompt}
-                                  onChange={(e) => setAiOptions({ ...aiOptions, userPrompt: e.target.value })}
-                                  className="bg-background resize-none text-sm"
-                                  rows={2}
-                                />
-                              </div>
-                              <Button
-                                onClick={handleAnalyzeExcel}
-                                disabled={excelAnalyzing}
-                                className="w-full gap-2 h-9"
-                                variant="outline"
-                              >
-                                {excelAnalyzing ? <Loader2 className="animate-spin" size={14} /> : <Sparkles size={14} />}
-                                {excelAnalyzing ? '分析中...' : 'AI 分析'}
-                              </Button>
-                              {excelParseResult?.success && (
-                                <Button
-                                  onClick={handleUseExcelData}
-                                  className="w-full gap-2 h-9"
-                                >
-                                  <CheckCircle2 size={14} />
-                                  使用此数据源
-                                </Button>
-                              )}
-                            </div>
-                          )}
-
-                          {/* Excel Analysis Result */}
-                          {excelAnalysis && (
-                            <div className="mt-4 p-4 bg-background rounded-xl max-h-60 overflow-y-auto">
-                              <div className="flex items-center gap-2 mb-3">
-                                <Sparkles size={16} className="text-primary" />
-                                <span className="font-semibold text-sm">AI 分析结果</span>
-                              </div>
-                              <Markdown content={excelAnalysis.analysis?.analysis || ''} className="text-sm" />
-                            </div>
-                          )}
-                        </div>
-                      )}
-                    </div>
-                  </div>
-
-                  {/* Step 3: Select Documents */}
-                  <div className="space-y-4">
-                    <h4 className="font-bold flex items-center gap-2 text-primary uppercase tracking-widest text-xs">
-                      <span className="w-5 h-5 rounded-full bg-primary text-white flex items-center justify-center text-[10px]">2</span>
-                      选择其他数据源文档
-                    </h4>
-                    {documents.filter(d => d.status === 'completed').length > 0 ? (
-                      <div className="space-y-2 max-h-40 overflow-y-auto pr-2 custom-scrollbar">
-                        {documents.filter(d => d.status === 'completed').map(doc => (
-                          <div
-                            key={doc.id}
-                            className={cn(
-                              "flex items-center gap-3 p-3 rounded-xl border transition-all cursor-pointer",
-                              selectedDocs.includes(doc.id) ? "border-primary/50 bg-primary/5 shadow-sm" : "border-border hover:bg-muted/30"
-                            )}
-                            onClick={() => {
-                              setSelectedDocs(prev =>
-                                prev.includes(doc.id) ? prev.filter(id => id !== doc.id) : [...prev, doc.id]
-                              );
-                            }}
-                          >
-                            <Checkbox checked={selectedDocs.includes(doc.id)} onCheckedChange={() => {}} />
-                            <div className="w-8 h-8 rounded-lg bg-blue-500/10 text-blue-500 flex items-center justify-center">
-                              <Zap size={16} />
-                            </div>
-                            <span className="font-semibold text-sm truncate">{doc.name}</span>
-                          </div>
-                        ))}
-                      </div>
-                    ) : (
-                      <div className="p-6 text-center bg-muted/30 rounded-xl border border-dashed text-xs italic text-muted-foreground">
-                        暂无其他已解析的文档
-                      </div>
-                    )}
-                  </div>
-                </div>
-              </ScrollArea>
-
-              <DialogFooter className="p-8 pt-4 bg-muted/20 border-t border-dashed">
-                <Button variant="outline" className="rounded-xl h-12 px-6" onClick={() => setOpenTaskDialog(false)}>取消</Button>
-                <Button
-                  className="rounded-xl h-12 px-8 shadow-lg shadow-primary/20 gap-2"
-                  onClick={handleCreateTask}
-                  disabled={creating || !selectedTemplate || (selectedDocs.length === 0 && !excelParseResult?.success)}
-                >
-                  {creating ? <RefreshCcw className="animate-spin h-5 w-5" /> : <Zap className="h-5 w-5 fill-current" />}
-                  <span>启动智能填表引擎</span>
-                </Button>
-              </DialogFooter>
-            </DialogContent>
-          </Dialog>
-        </div>
-      </section>
-
-      {/* Task List */}
-      <div className="grid grid-cols-1 md:grid-cols-2 lg:grid-cols-3 gap-6">
-        {loading ? (
-          Array.from({ length: 3 }).map((_, i) => (
-            <Skeleton key={i} className="h-48 w-full rounded-3xl bg-muted" />
-          ))
-        ) : tasks.length > 0 ? (
-          tasks.map((task) => (
-            <Card key={task.id} className="border-none shadow-md hover:shadow-xl transition-all group rounded-3xl overflow-hidden flex flex-col">
-              <div className="h-1.5 w-full" style={{ backgroundColor: task.status === 'completed' ? '#10b981' : task.status === 'failed' ? '#ef4444' : '#f59e0b' }} />
-              <CardHeader className="p-6 pb-2">
-                <div className="flex justify-between items-start mb-2">
-                  <div className="w-12 h-12 rounded-2xl bg-emerald-500/10 text-emerald-500 flex items-center justify-center shadow-inner group-hover:scale-110 transition-transform">
-                    <TableProperties size={24} />
-                  </div>
-                  <Badge className={cn("text-[10px] uppercase font-bold tracking-widest", getStatusColor(task.status))}>
-                    {task.status === 'completed' ? '已完成' : task.status === 'failed' ? '失败' : '执行中'}
-                  </Badge>
-                </div>
-                <CardTitle className="text-lg font-bold truncate group-hover:text-primary transition-colors">{task.templates?.name || '未知模板'}</CardTitle>
-                <CardDescription className="text-xs flex items-center gap-1 font-medium italic">
-                  <Clock size={12} /> {format(new Date(task.created_at!), 'yyyy/MM/dd HH:mm')}
-                </CardDescription>
-              </CardHeader>
-              <CardContent className="p-6 pt-2 flex-1">
-                <div className="space-y-4">
-                  <div className="flex flex-wrap gap-2">
-                    <Badge variant="outline" className="bg-muted/50 border-none text-[10px] font-bold">关联 {task.document_ids?.length} 份数据源</Badge>
-                  </div>
-                  {task.status === 'completed' && (
-                    <div className="p-3 bg-emerald-500/5 rounded-2xl border border-emerald-500/10 flex items-center gap-3">
-                      <CheckCircle2 className="text-emerald-500" size={18} />
-                      <span className="text-xs font-semibold text-emerald-700">内容已精准聚合，表格生成完毕</span>
-                    </div>
-                  )}
-                </div>
-              </CardContent>
-              <CardFooter className="p-6 pt-0">
-                <Button
-                  className="w-full rounded-2xl h-11 bg-primary group-hover:shadow-lg group-hover:shadow-primary/30 transition-all gap-2"
-                  disabled={task.status !== 'completed'}
-                  onClick={() => setViewingTask(task)}
-                >
-                  <Download size={18} />
-                  <span>下载汇总表格</span>
-                </Button>
-              </CardFooter>
-            </Card>
-          ))
-        ) : (
-          <div className="col-span-full py-24 flex flex-col items-center justify-center text-center space-y-6">
-            <div className="w-24 h-24 rounded-full bg-muted flex items-center justify-center text-muted-foreground/30 border-4 border-dashed">
-              <TableProperties size={48} />
-            </div>
-            <div className="space-y-2 max-w-sm">
-              <p className="text-2xl font-extrabold tracking-tight">暂无生成任务</p>
-              <p className="text-muted-foreground text-sm">上传模板后，您可以将多个文档的数据自动填充到汇总表格中。</p>
-            </div>
-            <Button className="rounded-xl h-12 px-8" onClick={() => setOpenTaskDialog(true)}>立即创建首个任务</Button>
-          </div>
-        )}
-      </div>
-
-      {/* Task Result View Modal */}
-      <Dialog open={!!viewingTask} onOpenChange={(open) => !open && setViewingTask(null)}>
-        <DialogContent className="max-w-4xl max-h-[90vh] flex flex-col p-0 overflow-hidden border-none shadow-2xl rounded-3xl">
-          <DialogHeader className="p-8 pb-4 bg-primary text-primary-foreground">
-            <div className="flex items-center gap-3 mb-2">
-              <FileCheck size={28} />
-              <DialogTitle className="text-2xl font-extrabold">表格生成结果预览</DialogTitle>
-            </div>
-            <DialogDescription className="text-primary-foreground/80 italic">
-              系统已根据 {viewingTask?.document_ids?.length} 份文档信息自动填充完毕。
-            </DialogDescription>
-          </DialogHeader>
-          <ScrollArea className="flex-1 p-8 bg-muted/10">
-            <div className="prose dark:prose-invert max-w-none">
-              <div className="bg-card p-8 rounded-2xl shadow-sm border min-h-[400px]">
-                <Badge variant="outline" className="mb-4">数据已脱敏</Badge>
-                <div className="whitespace-pre-wrap font-sans text-sm leading-relaxed">
-                  <h2 className="text-xl font-bold mb-4">汇总结果报告</h2>
-                  <p className="text-muted-foreground mb-6">以下是根据您上传的多个文档提取并生成的汇总信息：</p>
-
-                  <div className="p-4 bg-muted/30 rounded-xl border border-dashed border-primary/20 italic">
-                    正在从云端安全下载解析结果并渲染渲染视图...
-                  </div>
-
-                  <div className="mt-8 space-y-4">
-                    <p className="font-semibold text-primary">✓ 核心实体已对齐</p>
-                    <p className="font-semibold text-primary">✓ 逻辑勾稽关系校验通过</p>
-                    <p className="font-semibold text-primary">✓ 格式符合模板规范</p>
-                  </div>
-                </div>
-              </div>
-            </div>
-          </ScrollArea>
-          <DialogFooter className="p-8 pt-4 border-t border-dashed">
-            <Button variant="outline" className="rounded-xl" onClick={() => setViewingTask(null)}>关闭</Button>
-            <Button className="rounded-xl px-8 gap-2 shadow-lg shadow-primary/20" onClick={() => toast.success("正在导出文件...")}>
-              <Download size={18} />
-              导出为 {viewingTask?.templates?.type?.toUpperCase() || '文件'}
-            </Button>
-          </DialogFooter>
-        </DialogContent>
-      </Dialog>
-    </div>
-  );
-};
-
-export default FormFill;

frontend/src/pages/Login.tsx

@@ -1,184 +0,0 @@
-import React, { useState } from 'react';
-import { useNavigate, useLocation } from 'react-router-dom';
-import { useAuth } from '@/context/AuthContext';
-import { Button } from '@/components/ui/button';
-import { Input } from '@/components/ui/input';
-import { Label } from '@/components/ui/label';
-import { Card, CardContent, CardDescription, CardFooter, CardHeader, CardTitle } from '@/components/ui/card';
-import { Tabs, TabsContent, TabsList, TabsTrigger } from '@/components/ui/tabs';
-import { FileText, Lock, User, CheckCircle2, AlertCircle } from 'lucide-react';
-import { toast } from 'sonner';
-
-const Login: React.FC = () => {
-  const [username, setUsername] = useState('');
-  const [password, setPassword] = useState('');
-  const [loading, setLoading] = useState(false);
-  const { signIn, signUp } = useAuth();
-  const navigate = useNavigate();
-  const location = useLocation();
-
-  const handleLogin = async (e: React.FormEvent) => {
-    e.preventDefault();
-    if (!username || !password) return toast.error('请输入用户名和密码');
-    
-    setLoading(true);
-    try {
-      const email = `${username}@miaoda.com`;
-      const { error } = await signIn(email, password);
-      if (error) throw error;
-      toast.success('登录成功');
-      navigate('/');
-    } catch (err: any) {
-      toast.error(err.message || '登录失败');
-    } finally {
-      setLoading(false);
-    }
-  };
-
-  const handleSignUp = async (e: React.FormEvent) => {
-    e.preventDefault();
-    if (!username || !password) return toast.error('请输入用户名和密码');
-    
-    setLoading(true);
-    try {
-      const email = `${username}@miaoda.com`;
-      const { error } = await signUp(email, password);
-      if (error) throw error;
-      toast.success('注册成功，请登录');
-    } catch (err: any) {
-      toast.error(err.message || '注册失败');
-    } finally {
-      setLoading(false);
-    }
-  };
-
-  return (
-    <div className="min-h-screen flex items-center justify-center bg-[radial-gradient(ellipse_at_top_left,_var(--tw-gradient-stops))] from-primary/10 via-background to-background p-4 relative overflow-hidden">
-      {/* Decorative elements */}
-      <div className="absolute top-0 left-0 w-96 h-96 bg-primary/5 rounded-full blur-3xl -translate-x-1/2 -translate-y-1/2" />
-      <div className="absolute bottom-0 right-0 w-64 h-64 bg-primary/5 rounded-full blur-3xl translate-x-1/3 translate-y-1/3" />
-
-      <div className="w-full max-w-md space-y-8 relative animate-fade-in">
-        <div className="text-center space-y-2">
-          <div className="inline-flex items-center justify-center w-16 h-16 rounded-2xl bg-primary text-primary-foreground shadow-2xl shadow-primary/30 mb-4 animate-slide-in">
-            <FileText size={32} />
-          </div>
-          <h1 className="text-4xl font-extrabold tracking-tight gradient-text">智联文档</h1>
-          <p className="text-muted-foreground">多源数据融合与智能文档处理系统</p>
-        </div>
-
-        <Card className="border-border/50 shadow-2xl backdrop-blur-sm bg-card/95">
-          <Tabs defaultValue="login" className="w-full">
-            <TabsList className="grid w-full grid-cols-2 rounded-t-xl h-12 bg-muted/50 p-1">
-              <TabsTrigger value="login" className="rounded-lg data-[state=active]:bg-background data-[state=active]:shadow-sm">登录</TabsTrigger>
-              <TabsTrigger value="signup" className="rounded-lg data-[state=active]:bg-background data-[state=active]:shadow-sm">注册</TabsTrigger>
-            </TabsList>
-            
-            <TabsContent value="login">
-              <form onSubmit={handleLogin}>
-                <CardHeader>
-                  <CardTitle>欢迎回来</CardTitle>
-                  <CardDescription>使用您的账号登录智联文档系统</CardDescription>
-                </CardHeader>
-                <CardContent className="space-y-4">
-                  <div className="space-y-2">
-                    <Label htmlFor="username">用户名</Label>
-                    <div className="relative">
-                      <User className="absolute left-3 top-2.5 h-4 w-4 text-muted-foreground" />
-                      <Input 
-                        id="username" 
-                        placeholder="请输入用户名" 
-                        className="pl-9 bg-muted/30 border-none focus-visible:ring-primary"
-                        value={username} 
-                        onChange={(e) => setUsername(e.target.value)} 
-                      />
-                    </div>
-                  </div>
-                  <div className="space-y-2">
-                    <Label htmlFor="password">密码</Label>
-                    <div className="relative">
-                      <Lock className="absolute left-3 top-2.5 h-4 w-4 text-muted-foreground" />
-                      <Input 
-                        id="password" 
-                        type="password" 
-                        placeholder="请输入密码" 
-                        className="pl-9 bg-muted/30 border-none focus-visible:ring-primary"
-                        value={password} 
-                        onChange={(e) => setPassword(e.target.value)} 
-                      />
-                    </div>
-                  </div>
-                </CardContent>
-                <CardFooter>
-                  <Button className="w-full h-11 text-lg font-semibold rounded-xl" type="submit" disabled={loading}>
-                    {loading ? '登录中...' : '立即登录'}
-                  </Button>
-                </CardFooter>
-              </form>
-            </TabsContent>
-
-            <TabsContent value="signup">
-              <form onSubmit={handleSignUp}>
-                <CardHeader>
-                  <CardTitle>创建账号</CardTitle>
-                  <CardDescription>开启智能文档处理的新体验</CardDescription>
-                </CardHeader>
-                <CardContent className="space-y-4">
-                  <div className="space-y-2">
-                    <Label htmlFor="signup-username">用户名</Label>
-                    <div className="relative">
-                      <User className="absolute left-3 top-2.5 h-4 w-4 text-muted-foreground" />
-                      <Input 
-                        id="signup-username" 
-                        placeholder="仅字母、数字和下划线" 
-                        className="pl-9 bg-muted/30 border-none focus-visible:ring-primary"
-                        value={username} 
-                        onChange={(e) => setUsername(e.target.value)} 
-                      />
-                    </div>
-                  </div>
-                  <div className="space-y-2">
-                    <Label htmlFor="signup-password">密码</Label>
-                    <div className="relative">
-                      <Lock className="absolute left-3 top-2.5 h-4 w-4 text-muted-foreground" />
-                      <Input 
-                        id="signup-password" 
-                        type="password" 
-                        placeholder="不少于 6 位" 
-                        className="pl-9 bg-muted/30 border-none focus-visible:ring-primary"
-                        value={password} 
-                        onChange={(e) => setPassword(e.target.value)} 
-                      />
-                    </div>
-                  </div>
-                </CardContent>
-                <CardFooter>
-                  <Button className="w-full h-11 text-lg font-semibold rounded-xl" type="submit" disabled={loading}>
-                    {loading ? '注册中...' : '注册账号'}
-                  </Button>
-                </CardFooter>
-              </form>
-            </TabsContent>
-          </Tabs>
-        </Card>
-
-        <div className="grid grid-cols-2 gap-4 text-center text-xs text-muted-foreground">
-          <div className="flex flex-col items-center gap-1">
-            <CheckCircle2 size={16} className="text-primary" />
-            <span>智能解析</span>
-          </div>
-          <div className="flex flex-col items-center gap-1">
-            <CheckCircle2 size={16} className="text-primary" />
-            <span>极速填表</span>
-          </div>
-        </div>
-
-        <div className="text-center text-sm text-muted-foreground">
-          &copy; 2026 智联文档 | 多源数据融合系统
-        </div>
-      </div>
-    </div>
-  );
-};
-
-export default Login;

frontend/src/pages/SamplePage.tsx

@@ -1,16 +0,0 @@
-/**
- * Sample Page
- */
-
-import PageMeta from "../components/common/PageMeta";
-
-export default function SamplePage() {
-  return (
-    <>
-      <PageMeta title="Home" description="Home Page Introduction" />
-      <div>
-        <h3>This is a sample page</h3>
-      </div>
-    </>
-  );
-}

frontend/src/pages/TaskHistory.tsx

@@ -11,7 +11,8 @@ import {
   ChevronDown,
   ChevronUp,
   Trash2,
-  AlertCircle
+  AlertCircle,
+  HelpCircle
 } from 'lucide-react';
 import { Card, CardContent, CardHeader, CardTitle, CardDescription } from '@/components/ui/card';
 import { Button } from '@/components/ui/button';
@@ -24,9 +25,9 @@ import { Skeleton } from '@/components/ui/skeleton';
 
 type Task = {
   task_id: string;
-  status: 'pending' | 'processing' | 'success' | 'failure';
+  status: 'pending' | 'processing' | 'success' | 'failure' | 'unknown';
   created_at: string;
-  completed_at?: string;
+  updated_at?: string;
   message?: string;
   result?: any;
   error?: string;
@@ -38,54 +39,38 @@ const TaskHistory: React.FC = () => {
   const [loading, setLoading] = useState(true);
   const [expandedTask, setExpandedTask] = useState<string | null>(null);
 
-  // Mock data for demonstration
-  useEffect(() => {
-    // 模拟任务数据，实际应该从后端获取
-    setTasks([
-      {
-        task_id: 'task-001',
-        status: 'success',
-        created_at: new Date(Date.now() - 3600000).toISOString(),
-        completed_at: new Date(Date.now() - 3500000).toISOString(),
-        task_type: 'document_parse',
-        message: '文档解析完成',
-        result: {
-          doc_id: 'doc-001',
-          filename: 'report_q1_2026.docx',
-          extracted_fields: ['标题', '作者', '日期', '金额']
+  // 获取任务历史数据
+  const fetchTasks = async () => {
+    try {
+      setLoading(true);
+      const response = await backendApi.getTasks(50, 0);
+      if (response.success && response.tasks) {
+        // 转换后端数据格式为前端格式
+        const convertedTasks: Task[] = response.tasks.map((t: any) => ({
+          task_id: t.task_id,
+          status: t.status || 'unknown',
+          created_at: t.created_at || new Date().toISOString(),
+          updated_at: t.updated_at,
+          message: t.message || '',
+          result: t.result,
+          error: t.error,
+          task_type: t.task_type || 'document_parse'
+        }));
+        setTasks(convertedTasks);
+      } else {
+        setTasks([]);
       }
-      },
-      {
-        task_id: 'task-002',
-        status: 'success',
-        created_at: new Date(Date.now() - 7200000).toISOString(),
-        completed_at: new Date(Date.now() - 7100000).toISOString(),
-        task_type: 'excel_analysis',
-        message: 'Excel 分析完成',
-        result: {
-          filename: 'sales_data.xlsx',
-          row_count: 1250,
-          charts_generated: 3
-        }
-      },
-      {
-        task_id: 'task-003',
-        status: 'processing',
-        created_at: new Date(Date.now() - 600000).toISOString(),
-        task_type: 'template_fill',
-        message: '正在填充表格...'
-      },
-      {
-        task_id: 'task-004',
-        status: 'failure',
-        created_at: new Date(Date.now() - 86400000).toISOString(),
-        completed_at: new Date(Date.now() - 86390000).toISOString(),
-        task_type: 'document_parse',
-        message: '解析失败',
-        error: '文件格式不支持或文件已损坏'
-      }
-    ]);
+    } catch (error) {
+      console.error('获取任务列表失败:', error);
+      toast.error('获取任务列表失败');
+      setTasks([]);
+    } finally {
       setLoading(false);
+    }
+  };
+
+  useEffect(() => {
+    fetchTasks();
   }, []);
 
   const getStatusBadge = (status: string) => {
@@ -96,6 +81,8 @@ const TaskHistory: React.FC = () => {
         return <Badge className="bg-destructive text-white text-[10px]"><XCircle size={12} className="mr-1" />失败</Badge>;
       case 'processing':
         return <Badge className="bg-amber-500 text-white text-[10px]"><Loader2 size={12} className="mr-1 animate-spin" />处理中</Badge>;
+      case 'unknown':
+        return <Badge className="bg-gray-500 text-white text-[10px]"><HelpCircle size={12} className="mr-1" />未知</Badge>;
       default:
         return <Badge className="bg-gray-500 text-white text-[10px]"><Clock size={12} className="mr-1" />等待</Badge>;
     }
@@ -133,15 +120,22 @@ const TaskHistory: React.FC = () => {
   };
 
   const handleDelete = async (taskId: string) => {
+    try {
+      await backendApi.deleteTask(taskId);
       setTasks(prev => prev.filter(t => t.task_id !== taskId));
       toast.success('任务已删除');
+    } catch (error) {
+      console.error('删除任务失败:', error);
+      toast.error('删除任务失败');
+    }
   };
 
   const stats = {
     total: tasks.length,
     success: tasks.filter(t => t.status === 'success').length,
     processing: tasks.filter(t => t.status === 'processing').length,
-    failure: tasks.filter(t => t.status === 'failure').length
+    failure: tasks.filter(t => t.status === 'failure').length,
+    unknown: tasks.filter(t => t.status === 'unknown').length
   };
 
   return (
@@ -151,7 +145,7 @@ const TaskHistory: React.FC = () => {
           <h1 className="text-3xl font-extrabold tracking-tight">任务历史</h1>
           <p className="text-muted-foreground">查看和管理您所有的文档处理任务记录</p>
         </div>
-        <Button variant="outline" className="rounded-xl gap-2" onClick={() => window.location.reload()}>
+        <Button variant="outline" className="rounded-xl gap-2" onClick={() => fetchTasks()}>
           <RefreshCcw size={18} />
           <span>刷新</span>
         </Button>
@@ -194,7 +188,8 @@ const TaskHistory: React.FC = () => {
                     "w-12 h-12 rounded-xl flex items-center justify-center shrink-0",
                     task.status === 'success' ? "bg-emerald-500/10 text-emerald-500" :
                     task.status === 'failure' ? "bg-destructive/10 text-destructive" :
-                    "bg-amber-500/10 text-amber-500"
+                    task.status === 'processing' ? "bg-amber-500/10 text-amber-500" :
+                    "bg-gray-500/10 text-gray-500"
                   )}>
                     {task.status === 'processing' ? (
                       <Loader2 size={24} className="animate-spin" />
@@ -212,16 +207,16 @@ const TaskHistory: React.FC = () => {
                       </Badge>
                     </div>
                     <p className="text-sm text-muted-foreground">
-                      {task.message || '任务执行中...'}
+                      {task.message || (task.status === 'unknown' ? '无法获取状态' : '任务执行中...')}
                     </p>
                     <div className="flex items-center gap-4 text-xs text-muted-foreground">
                       <span className="flex items-center gap-1">
                         <Clock size={12} />
-                        {format(new Date(task.created_at), 'yyyy-MM-dd HH:mm:ss')}
+                        {task.created_at ? format(new Date(task.created_at), 'yyyy-MM-dd HH:mm:ss') : '时间未知'}
                       </span>
-                      {task.completed_at && (
+                      {task.updated_at && task.status !== 'processing' && (
                         <span>
-                          耗时: {Math.round((new Date(task.completed_at).getTime() - new Date(task.created_at).getTime()) / 1000)} 秒
+                          更新: {format(new Date(task.updated_at), 'HH:mm:ss')}
                         </span>
                       )}
                     </div>

frontend/src/pages/TemplateFill.tsx

@@ -626,6 +626,16 @@ const TemplateFill: React.FC = () => {
                         <div className="text-muted-foreground text-xs mt-1">
                           来源: {detail.source} | 置信度: {detail.confidence ? (detail.confidence * 100).toFixed(0) + '%' : 'N/A'}
                         </div>
+                        {detail.warning && (
+                          <div className="mt-2 p-2 bg-yellow-50 border border-yellow-200 rounded-lg text-yellow-700 text-xs">
+                            ⚠️ {detail.warning}
+                          </div>
+                        )}
+                        {detail.values && detail.values.length > 1 && !detail.warning && (
+                          <div className="mt-2 text-xs text-muted-foreground">
+                            多值: {detail.values.join(', ')}
+                          </div>
+                        )}
                       </div>
                     </div>
                   ))}

比赛备赛规划.md

@@ -50,18 +50,18 @@
 | `prompt_service.py` | ✅ 已完成 | Prompt 模板管理 |
 | `text_analysis_service.py` | ✅ 已完成 | 文本分析 |
 | `chart_generator_service.py` | ✅ 已完成 | 图表生成服务 |
-| `template_fill_service.py` | ✅ 已完成 | 模板填写服务，支持直接读取源文档进行填表 |
+| `template_fill_service.py` | ✅ 已完成 | 模板填写服务，支持多行提取、直接从结构化数据提取、JSON容错、Word文档表格处理 |
 
 ### 2.2 API 接口 (`backend/app/api/endpoints/`)
 
 | 接口文件 | 路由 | 功能状态 |
 |----------|------|----------|
-| `upload.py` | `/api/v1/upload/excel` | ✅ Excel 文件上传与解析 |
+| `upload.py` | `/api/v1/upload/document` | ✅ 文档上传与解析 |
 | `documents.py` | `/api/v1/documents/*` | ✅ 文档管理（列表、删除、搜索） |
 | `ai_analyze.py` | `/api/v1/analyze/*` | ✅ AI 分析（Excel、Markdown、流式） |
 | `rag.py` | `/api/v1/rag/*` | ⚠️ RAG 检索（当前返回空） |
 | `tasks.py` | `/api/v1/tasks/*` | ✅ 异步任务状态查询 |
-| `templates.py` | `/api/v1/templates/*` | ✅ 模板管理 (含 Word 导出) |
+| `templates.py` | `/api/v1/templates/*` | ✅ 模板管理（含多行导出、Word导出、Word结构化字段解析） |
 | `visualization.py` | `/api/v1/visualization/*` | ✅ 可视化图表 |
 | `health.py` | `/api/v1/health` | ✅ 健康检查 |
 
@@ -70,71 +70,67 @@
 | 页面文件 | 功能 | 状态 |
 |----------|------|------|
 | `Documents.tsx` | 主文档管理页面 | ✅ 已完成 |
+| `TemplateFill.tsx` | 智能填表页面 | ✅ 已完成 |
 | `ExcelParse.tsx` | Excel 解析页面 | ✅ 已完成 |
 
 ### 2.4 文档解析能力
 
 | 格式 | 解析状态 | 说明 |
 |------|----------|------|
-| Excel (.xlsx/.xls) | ✅ 已完成 | pandas + XML 回退解析 |
+| Excel (.xlsx/.xls) | ✅ 已完成 | pandas + XML 回退解析，支持多sheet |
 | Markdown (.md) | ✅ 已完成 | 正则 + AI 分章节 |
 | Word (.docx) | ✅ 已完成 | python-docx 解析，支持表格提取和字段识别 |
 | Text (.txt) | ✅ 已完成 | chardet 编码检测，支持文本清洗和结构化提取 |
 
 ---
 
-## 三、待完成功能（核心缺块）
+## 三、核心功能实现详情
 
-### 3.1 模板填写模块（最优先）
-
-**当前状态**：✅ 已完成
+### 3.1 模板填写模块（✅ 已完成）
 
+**核心流程**：
 ```
-用户上传模板表格(Word/Excel)
+上传模板表格(Word/Excel)
          ↓
 解析模板，提取需要填写的字段和提示词
          ↓
-根据模板指定的源文档列表读取源数据
+根据源文档ID列表读取源数据（MongoDB或文件）
          ↓
-AI 根据字段提示词从源数据中提取信息
+优先从结构化数据直接提取（Excel rows）
          ↓
-将提取的数据填入模板对应位置
+无法直接提取时使用 LLM 从文本中提取
          ↓
-返回填写完成的表格
+将提取的数据填入原始模板对应位置（保持模板格式）
+         ↓
+导出填写完成的表格（Excel/Word）
 ```
 
-**已完成实现**：
-- [x] `template_fill_service.py` - 模板填写核心服务
-- [x] Word 模板解析 (`docx_parser.py` - parse_tables_for_template, extract_template_fields_from_docx)
-- [x] Text 模板解析 (`txt_parser.py` - 已完成)
-- [x] 模板字段识别与提示词提取
-- [x] 多文档数据聚合与冲突处理
-- [x] 结果导出为 Word/Excel
+**关键特性**：
+- **原始模板填充**：直接打开原始模板文件，填充数据到原表格/单元格
+- **多行数据支持**：每个字段可提取多个值，导出时自动扩展行数
+- **结构化数据优先**：直接从 Excel rows 提取，无需 LLM
+- **JSON 容错**：支持 LLM 返回的损坏/截断 JSON
+- **Markdown 清理**：自动清理 LLM 返回的 markdown 格式
 
-### 3.2 Word 文档解析
-
-**当前状态**：✅ 已完成
+### 3.2 Word 文档解析（✅ 已完成）
 
 **已实现功能**：
-- [x] `docx_parser.py` - Word 文档解析器
-- [x] 提取段落文本
-- [x] 提取表格内容
-- [x] 提取关键信息（标题、列表等）
-- [x] 表格模板字段提取 (`parse_tables_for_template`, `extract_template_fields_from_docx`)
-- [x] 字段类型推断 (`_infer_field_type_from_hint`)
+- `docx_parser.py` - Word 文档解析器
+- 提取段落文本
+- 提取表格内容（支持比赛表格格式：字段名 | 提示词 | 填写值）
+- `parse_tables_for_template()` - 解析表格模板，提取字段
+- `extract_template_fields_from_docx()` - 提取模板字段定义
+- `_infer_field_type_from_hint()` - 从提示词推断字段类型
+- **API 端点**：`/api/v1/templates/parse-word-structure` - 上传 Word 文档，提取结构化字段并存入 MongoDB
+- **API 端点**：`/api/v1/templates/word-fields/{doc_id}` - 获取已存文档的模板字段信息
 
-### 3.3 Text 文档解析
-
-**当前状态**：✅ 已完成
+### 3.3 Text 文档解析（✅ 已完成）
 
 **已实现功能**：
-- [x] `txt_parser.py` - 文本文件解析器
-- [x] 编码自动检测 (chardet)
-- [x] 文本清洗
-
-### 3.4 文档模板匹配（已有框架）
-
-根据 Q&A，模板已指定数据文件，不需要算法匹配。当前已有上传功能，需确认模板与数据文件的关联逻辑是否完善。
+- `txt_parser.py` - 文本文件解析器
+- 编码自动检测 (chardet)
+- 文本清洗（去除控制字符、规范化空白）
+- 结构化数据提取（邮箱、URL、电话、日期、金额）
 
 ---
 
@@ -192,20 +188,20 @@ docs/test/
 
 ## 六、工作计划（建议）
 
-### 第一优先级：模板填写核心功能
-- 完成 Word 文档解析
-- 完成模板填写服务
-- 端到端测试验证
+### 第一优先级：端到端测试
+- 使用真实测试数据进行准确率测试
+- 验证多行数据导出是否正确
+- 测试 Word 模板解析是否正常
 
 ### 第二优先级：Demo 打包与文档
 - 制作项目演示 PPT
 - 录制演示视频
 - 完善 README 部署文档
 
-### 第三优先级：测试优化
-- 使用真实测试数据进行准确率测试
+### 第三优先级：优化
 - 优化响应时间
 - 完善错误处理
+- 增加更多测试用例
 
 ---
 
@@ -215,29 +211,32 @@ docs/test/
 2. **数据库**：不强制要求数据库存储，可跳过
 3. **部署**：本地部署即可，不需要公网服务器
 4. **评测数据**：初赛仅使用目前提供的数据
-5. **RAG 功能**：当前已临时禁用，不影响核心评测功能
+5. **RAG 功能**：当前已临时禁用，不影响核心评测功能（因为使用直接文件读取）
 
 ---
 
-*文档版本: v1.1*
-*最后更新: 2026-04-08*
+*文档版本: v1.5*
+*最后更新: 2026-04-09*
 
 ---
 
 ## 八、技术实现细节
 
-### 8.1 模板填表流程（已实现）
+### 8.1 模板填表流程
 
 #### 流程图
 ```
 ┌─────────────┐     ┌─────────────┐     ┌─────────────┐
-│ 上传模板    │ ──► │ 选择数据源   │ ──► │ AI 智能填表  │
+│ 上传模板    │ ──► │ 选择数据源   │ ──► │ 智能填表    │
 └─────────────┘     └─────────────┘     └─────────────┘
                                               │
-                                               ▼
-                                        ┌─────────────┐
-                                        │ 导出结果    │
-                                        └─────────────┘
+                    ┌─────────────────────────┼─────────────────────────┐
+                    │                         │                         │
+                    ▼                         ▼                         ▼
+           ┌───────────────┐         ┌───────────────┐         ┌───────────────┐
+           │ 结构化数据提取 │         │   LLM 提取    │         │   导出结果    │
+           │ (直接读rows)  │         │ (文本理解)    │         │ (Excel/Word) │
+           └───────────────┘         └───────────────┘         └───────────────┘
 ```
 
 #### 核心组件
@@ -247,8 +246,10 @@ docs/test/
 | 模板上传 | `templates.py` `/templates/upload` | 接收模板文件，提取字段 |
 | 字段提取 | `template_fill_service.py` | 从 Word/Excel 表格提取字段定义 |
 | 文档解析 | `docx_parser.py`, `xlsx_parser.py`, `txt_parser.py` | 解析源文档内容 |
-| 智能填表 | `template_fill_service.py` `fill_template()` | 使用 LLM 从源文档提取信息 |
-| 结果导出 | `templates.py` `/templates/export` | 导出为 Excel 或 Word |
+| 智能填表 | `template_fill_service.py` `fill_template()` | 结构化提取 + LLM 提取 |
+| 多行支持 | `template_fill_service.py` `FillResult` | values 数组支持 |
+| JSON 容错 | `template_fill_service.py` `_fix_json()` | 修复损坏的 JSON |
+| 结果导出 | `templates.py` `/templates/export` | 多行 Excel + Word 导出 |
 
 ### 8.2 源文档加载方式
 
@@ -268,7 +269,9 @@ docs/test/
 
 ```python
 # 提取表格模板字段
-fields = docx_parser.extract_template_fields_from_docx(file_path)
+from docx_parser import DocxParser
+parser = DocxParser()
+fields = parser.extract_template_fields_from_docx(file_path)
 
 # 返回格式
 # [
@@ -295,6 +298,24 @@ fields = docx_parser.extract_template_fields_from_docx(file_path)
 
 ### 8.5 API 接口
 
+#### POST `/api/v1/templates/upload`
+
+上传模板文件，提取字段定义。
+
+**响应**：
+```json
+{
+  "success": true,
+  "template_id": "/path/to/saved/template.docx",
+  "filename": "模板.docx",
+  "file_type": "docx",
+  "fields": [
+    {"cell": "A1", "name": "姓名", "field_type": "text", "required": true, "hint": "提取人员姓名"}
+  ],
+  "field_count": 1
+}
+```
+
 #### POST `/api/v1/templates/fill`
 
 填写请求：
@@ -306,35 +327,232 @@ fields = docx_parser.extract_template_fields_from_docx(file_path)
   ],
   "source_doc_ids": ["mongodb_doc_id_1", "mongodb_doc_id_2"],
   "source_file_paths": [],
-  "user_hint": "请从合同文档中提取"
+  "user_hint": "请从xxx文档中提取"
 }
 ```
 
-响应：
+**响应（含多行支持）**：
 ```json
 {
   "success": true,
-  "filled_data": {"姓名": "张三"},
+  "filled_data": {
+    "姓名": ["张三", "李四", "王五"],
+    "年龄": ["25", "30", "28"]
+  },
   "fill_details": [
     {
       "field": "姓名",
       "cell": "A1",
+      "values": ["张三", "李四", "王五"],
       "value": "张三",
-      "source": "来自：合同文档.docx",
-      "confidence": 0.95
+      "source": "结构化数据直接提取",
+      "confidence": 1.0
     }
   ],
-  "source_doc_count": 2
+  "source_doc_count": 2,
+  "max_rows": 3
 }
 ```
 
 #### POST `/api/v1/templates/export`
 
-导出请求：
+导出请求（创建新文件）：
 ```json
 {
   "template_id": "模板ID",
-  "filled_data": {"姓名": "张三", "金额": "10000"},
-  "format": "xlsx"  // 或 "docx"
+  "filled_data": {"姓名": ["张三", "李四"], "金额": ["10000", "20000"]},
+  "format": "xlsx"
 }
 ```
+
+#### POST `/api/v1/templates/fill-and-export`
+
+**填充原始模板并导出**（推荐用于比赛）
+
+直接打开原始模板文件，将数据填入模板的表格/单元格中，然后导出。**保持原始模板格式不变**。
+
+**请求**：
+```json
+{
+  "template_path": "/path/to/original/template.docx",
+  "filled_data": {
+    "姓名": ["张三", "李四", "王五"],
+    "年龄": ["25", "30", "28"]
+  },
+  "format": "docx"
+}
+```
+
+**响应**：填充后的 Word/Excel 文件（文件流）
+
+**特点**：
+- 打开原始模板文件
+- 根据表头行匹配字段名到列索引
+- 将数据填入对应列的单元格
+- 多行数据自动扩展表格行数
+- 保持原始模板格式和样式
+
+#### POST `/api/v1/templates/parse-word-structure`
+
+**上传 Word 文档并提取结构化字段**（比赛专用）
+
+上传 Word 文档，从表格模板中提取字段定义（字段名、提示词、字段类型）并存入 MongoDB。
+
+**请求**：multipart/form-data
+- file: Word 文件
+
+**响应**：
+```json
+{
+  "success": true,
+  "doc_id": "mongodb_doc_id",
+  "filename": "模板.docx",
+  "file_path": "/path/to/saved/template.docx",
+  "field_count": 5,
+  "fields": [
+    {
+      "cell": "T0R1",
+      "name": "字段名",
+      "hint": "提示词",
+      "field_type": "text",
+      "required": true
+    }
+  ],
+  "tables": [...],
+  "metadata": {
+    "paragraph_count": 10,
+    "table_count": 1,
+    "word_count": 500,
+    "has_tables": true
+  }
+}
+```
+
+#### GET `/api/v1/templates/word-fields/{doc_id}`
+
+**获取 Word 文档模板字段信息**
+
+根据 doc_id 获取已上传的 Word 文档的模板字段信息。
+
+**响应**：
+```json
+{
+  "success": true,
+  "doc_id": "mongodb_doc_id",
+  "filename": "模板.docx",
+  "fields": [...],
+  "tables": [...],
+  "field_count": 5,
+  "metadata": {...}
+}
+```
+
+### 8.6 多行数据处理
+
+**FillResult 数据结构**：
+```python
+@dataclass
+class FillResult:
+    field: str
+    values: List[Any] = None  # 支持多个值（数组）
+    value: Any = ""           # 保留兼容（第一个值）
+    source: str = ""          # 来源文档
+    confidence: float = 1.0   # 置信度
+```
+
+**导出逻辑**：
+- 计算所有字段的最大行数
+- 遍历每一行，取对应索引的值
+- 不足的行填空字符串
+
+### 8.7 JSON 容错处理
+
+当 LLM 返回的 JSON 损坏或被截断时，系统会：
+
+1. 清理 markdown 代码块标记（```json, ```）
+2. 尝试配对括号找到完整的 JSON
+3. 移除末尾多余的逗号
+4. 使用正则表达式提取 values 数组
+5. 备选方案：直接提取所有引号内的字符串
+
+### 8.8 结构化数据优先提取
+
+对于 Excel 等有 `rows` 结构的文档，系统会：
+
+1. 直接从 `structured_data.rows` 中查找匹配列
+2. 使用模糊匹配（字段名包含或被包含）
+3. 提取该列的所有行值
+4. 无需调用 LLM，速度更快，准确率更高
+
+```python
+# 内部逻辑
+if structured.get("rows"):
+    columns = structured.get("columns", [])
+    values = _extract_column_values(rows, columns, field_name)
+```
+
+---
+
+## 九、依赖说明
+
+### Python 依赖
+
+```
+# requirements.txt 中需要包含
+fastapi>=0.104.0
+uvicorn>=0.24.0
+motor>=3.3.0          # MongoDB 异步驱动
+sqlalchemy>=2.0.0     # MySQL ORM
+pandas>=2.0.0         # Excel 处理
+openpyxl>=3.1.0       # Excel 写入
+python-docx>=0.8.0    # Word 处理
+chardet>=4.0.0        # 编码检测
+httpx>=0.25.0         # HTTP 客户端
+```
+
+### 前端依赖
+
+```
+# package.json 中需要包含
+react>=18.0.0
+react-dropzone>=14.0.0
+lucide-react>=0.300.0
+sonner>=1.0.0         # toast 通知
+```
+
+---
+
+## 十、启动说明
+
+### 后端启动
+
+```bash
+cd backend
+.\venv\Scripts\Activate.ps1  # 或 Activate.bat
+pip install -r requirements.txt  # 确保依赖完整
+.\venv\Scripts\python.exe -m uvicorn app.main:app --host 127.0.0.1 --port 8000 --reload
+```
+
+### 前端启动
+
+```bash
+cd frontend
+npm install
+npm run dev
+```
+
+### 环境变量
+
+在 `backend/.env` 中配置：
+```
+MONGODB_URL=mongodb://localhost:27017
+MONGODB_DB_NAME=document_system
+MYSQL_HOST=localhost
+MYSQL_PORT=3306
+MYSQL_USER=root
+MYSQL_PASSWORD=your_password
+MYSQL_DATABASE=document_system
+LLM_API_KEY=your_api_key
+LLM_BASE_URL=https://api.minimax.chat
+LLM_MODEL_NAME=MiniMax-Text-01
+```