feat: 实现智能指令的格式转换和文档编辑功能

主要更新：
- 新增 transform 意图：支持 Word/Excel/Markdown 格式互转
- 新增 edit 意图：使用 LLM 润色编辑文档内容
- 智能指令接口增加异步执行模式（async_execute 参数）
- 修复 Word 模板导出文档损坏问题（改用临时文件方式）
- 优化 intent_parser 增加 transform/edit 关键词识别

新增文件：
- app/api/endpoints/instruction.py: 智能指令 API 端点
- app/services/multi_doc_reasoning_service.py: 多文档推理服务

其他优化：
- RAG 服务混合搜索（BM25 + 向量）融合
- 模板填充服务表头匹配增强
- Word AI 解析服务返回结构完善
- 前端 InstructionChat 组件对接真实 API

backend/app/instruction/__init__.py

@@ -1,15 +1,14 @@
 """
 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
+支持文档智能操作交互，包括意图解析和指令执行
 """
-from .intent_parser import IntentParser, DefaultIntentParser
-from .executor import InstructionExecutor, DefaultInstructionExecutor
+from .intent_parser import IntentParser, intent_parser
+from .executor import InstructionExecutor, instruction_executor
 
 __all__ = [
     "IntentParser",
-    "DefaultIntentParser",
+    "intent_parser",
     "InstructionExecutor",
-    "DefaultInstructionExecutor",
+    "instruction_executor",
 ]

backend/app/instruction/executor.py

@@ -2,34 +2,571 @@
 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
+import logging
+import json
+from typing import Any, Dict, List, Optional
+
+from app.services.template_fill_service import template_fill_service
+from app.services.rag_service import rag_service
+from app.services.markdown_ai_service import markdown_ai_service
+from app.core.database import mongodb
+
+logger = logging.getLogger(__name__)
 
 
-class InstructionExecutor(ABC):
-    """指令执行器抽象基类"""
+class InstructionExecutor:
+    """指令执行器"""
 
-    @abstractmethod
-    async def execute(self, instruction: str, context: Dict[str, Any]) -> Dict[str, Any]:
+    def __init__(self):
+        self.intent_parser = None  # 将通过 set_intent_parser 设置
+
+    def set_intent_parser(self, intent_parser):
+        """设置意图解析器"""
+        self.intent_parser = intent_parser
+
+    async def execute(self, instruction: str, context: Dict[str, Any] = None) -> Dict[str, Any]:
         """
         执行指令
 
         Args:
-            instruction: 解析后的指令
-            context: 执行上下文
+            instruction: 自然语言指令
+            context: 执行上下文（包含文档信息等）
 
         Returns:
             执行结果
         """
-        pass
+        if self.intent_parser is None:
+            from app.instruction.intent_parser import intent_parser
+            self.intent_parser = intent_parser
+
+        context = context or {}
+
+        # 解析意图
+        intent, params = await self.intent_parser.parse(instruction)
+
+        # 根据意图类型执行相应操作
+        if intent == "extract":
+            return await self._execute_extract(params, context)
+        elif intent == "fill_table":
+            return await self._execute_fill_table(params, context)
+        elif intent == "summarize":
+            return await self._execute_summarize(params, context)
+        elif intent == "question":
+            return await self._execute_question(params, context)
+        elif intent == "search":
+            return await self._execute_search(params, context)
+        elif intent == "compare":
+            return await self._execute_compare(params, context)
+        elif intent == "edit":
+            return await self._execute_edit(params, context)
+        elif intent == "transform":
+            return await self._execute_transform(params, context)
+        else:
+            return {
+                "success": False,
+                "error": f"未知意图类型: {intent}",
+                "message": "无法理解该指令，请尝试更明确的描述"
+            }
+
+    async def _execute_extract(self, params: Dict[str, Any], context: Dict[str, Any]) -> Dict[str, Any]:
+        """执行信息提取"""
+        try:
+            target_fields = params.get("field_refs", [])
+            doc_ids = params.get("document_refs", [])
+
+            if not target_fields:
+                return {
+                    "success": False,
+                    "error": "未指定要提取的字段",
+                    "message": "请明确说明要提取哪些字段，如：'提取医院数量和床位数'"
+                }
+
+            # 如果指定了文档，验证文档存在
+            if doc_ids and "all_docs" not in doc_ids:
+                valid_docs = []
+                for doc_ref in doc_ids:
+                    doc_id = doc_ref.replace("doc_", "")
+                    doc = await mongodb.get_document(doc_id)
+                    if doc:
+                        valid_docs.append(doc)
+                if not valid_docs:
+                    return {
+                        "success": False,
+                        "error": "指定的文档不存在",
+                        "message": "请检查文档编号是否正确"
+                    }
+                context["source_docs"] = valid_docs
+
+            # 构建字段列表
+            fields = []
+            for i, field_name in enumerate(target_fields):
+                fields.append({
+                    "name": field_name,
+                    "cell": f"A{i+1}",
+                    "field_type": "text",
+                    "required": False
+                })
+
+            # 调用填表服务
+            result = await template_fill_service.fill_template(
+                template_fields=fields,
+                source_doc_ids=[doc.get("_id") for doc in context.get("source_docs", [])] if context.get("source_docs") else None,
+                user_hint=f"请提取字段: {', '.join(target_fields)}"
+            )
+
+            return {
+                "success": True,
+                "intent": "extract",
+                "extracted_data": result.get("filled_data", {}),
+                "fields": target_fields,
+                "message": f"成功提取 {len(result.get('filled_data', {}))} 个字段"
+            }
+
+        except Exception as e:
+            logger.error(f"提取执行失败: {e}")
+            return {
+                "success": False,
+                "error": str(e),
+                "message": f"提取失败: {str(e)}"
+            }
+
+    async def _execute_fill_table(self, params: Dict[str, Any], context: Dict[str, Any]) -> Dict[str, Any]:
+        """执行填表操作"""
+        try:
+            template_file = context.get("template_file")
+            if not template_file:
+                return {
+                    "success": False,
+                    "error": "未提供表格模板",
+                    "message": "请先上传要填写的表格模板"
+                }
+
+            # 获取源文档
+            source_docs = context.get("source_docs", [])
+            source_doc_ids = [doc.get("_id") for doc in source_docs if doc.get("_id")]
+
+            # 获取字段
+            fields = context.get("template_fields", [])
+
+            # 调用填表服务
+            result = await template_fill_service.fill_template(
+                template_fields=fields,
+                source_doc_ids=source_doc_ids if source_doc_ids else None,
+                source_file_paths=context.get("source_file_paths"),
+                user_hint=params.get("user_hint"),
+                template_id=template_file if isinstance(template_file, str) else None,
+                template_file_type=params.get("template", {}).get("type", "xlsx")
+            )
+
+            return {
+                "success": True,
+                "intent": "fill_table",
+                "result": result,
+                "message": f"填表完成，成功填写 {len(result.get('filled_data', {}))} 个字段"
+            }
+
+        except Exception as e:
+            logger.error(f"填表执行失败: {e}")
+            return {
+                "success": False,
+                "error": str(e),
+                "message": f"填表失败: {str(e)}"
+            }
+
+    async def _execute_summarize(self, params: Dict[str, Any], context: Dict[str, Any]) -> Dict[str, Any]:
+        """执行摘要总结"""
+        try:
+            docs = context.get("source_docs", [])
+            if not docs:
+                return {
+                    "success": False,
+                    "error": "没有可用的文档",
+                    "message": "请先上传要总结的文档"
+                }
+
+            summaries = []
+            for doc in docs[:5]:  # 最多处理5个文档
+                content = doc.get("content", "")[:5000]  # 限制内容长度
+                if content:
+                    summaries.append({
+                        "filename": doc.get("metadata", {}).get("original_filename", "未知"),
+                        "content_preview": content[:500] + "..." if len(content) > 500 else content
+                    })
+
+            return {
+                "success": True,
+                "intent": "summarize",
+                "summaries": summaries,
+                "message": f"找到 {len(summaries)} 个文档可供参考"
+            }
+
+        except Exception as e:
+            logger.error(f"摘要执行失败: {e}")
+            return {
+                "success": False,
+                "error": str(e),
+                "message": f"摘要生成失败: {str(e)}"
+            }
+
+    async def _execute_question(self, params: Dict[str, Any], context: Dict[str, Any]) -> Dict[str, Any]:
+        """执行问答"""
+        try:
+            question = params.get("question", "")
+            if not question:
+                return {
+                    "success": False,
+                    "error": "未提供问题",
+                    "message": "请输入要回答的问题"
+                }
+
+            # 使用 RAG 检索相关文档
+            docs = context.get("source_docs", [])
+            rag_results = []
+
+            for doc in docs:
+                doc_id = doc.get("_id", "")
+                if doc_id:
+                    results = rag_service.retrieve_by_doc_id(doc_id, top_k=3)
+                    rag_results.extend(results)
+
+            # 构建上下文
+            context_text = "\n\n".join([
+                r.get("content", "") for r in rag_results[:5]
+            ]) if rag_results else ""
+
+            # 如果没有 RAG 结果，使用文档内容
+            if not context_text:
+                context_text = "\n\n".join([
+                    doc.get("content", "")[:3000] for doc in docs[:3] if doc.get("content")
+                ])
+
+            return {
+                "success": True,
+                "intent": "question",
+                "question": question,
+                "context_preview": context_text[:500] + "..." if len(context_text) > 500 else context_text,
+                "message": "已找到相关上下文，可进行问答"
+            }
+
+        except Exception as e:
+            logger.error(f"问答执行失败: {e}")
+            return {
+                "success": False,
+                "error": str(e),
+                "message": f"问答处理失败: {str(e)}"
+            }
+
+    async def _execute_search(self, params: Dict[str, Any], context: Dict[str, Any]) -> Dict[str, Any]:
+        """执行搜索"""
+        try:
+            field_refs = params.get("field_refs", [])
+            query = " ".join(field_refs) if field_refs else params.get("question", "")
+
+            if not query:
+                return {
+                    "success": False,
+                    "error": "未提供搜索关键词",
+                    "message": "请输入要搜索的关键词"
+                }
+
+            # 使用 RAG 检索
+            results = rag_service.retrieve(query, top_k=10, min_score=0.3)
+
+            return {
+                "success": True,
+                "intent": "search",
+                "query": query,
+                "results": [
+                    {
+                        "content": r.get("content", "")[:200],
+                        "score": r.get("score", 0),
+                        "doc_id": r.get("doc_id", "")
+                    }
+                    for r in results[:10]
+                ],
+                "message": f"找到 {len(results)} 条相关结果"
+            }
+
+        except Exception as e:
+            logger.error(f"搜索执行失败: {e}")
+            return {
+                "success": False,
+                "error": str(e),
+                "message": f"搜索失败: {str(e)}"
+            }
+
+    async def _execute_compare(self, params: Dict[str, Any], context: Dict[str, Any]) -> Dict[str, Any]:
+        """执行对比分析"""
+        try:
+            docs = context.get("source_docs", [])
+            if len(docs) < 2:
+                return {
+                    "success": False,
+                    "error": "对比需要至少2个文档",
+                    "message": "请上传至少2个文档进行对比"
+                }
+
+            # 提取文档基本信息
+            comparison = []
+            for i, doc in enumerate(docs[:5]):
+                comparison.append({
+                    "index": i + 1,
+                    "filename": doc.get("metadata", {}).get("original_filename", "未知"),
+                    "doc_type": doc.get("doc_type", "未知"),
+                    "content_length": len(doc.get("content", "")),
+                    "has_tables": bool(doc.get("structured_data", {}).get("tables")),
+                })
+
+            return {
+                "success": True,
+                "intent": "compare",
+                "comparison": comparison,
+                "message": f"对比了 {len(comparison)} 个文档的基本信息"
+            }
+
+        except Exception as e:
+            logger.error(f"对比执行失败: {e}")
+            return {
+                "success": False,
+                "error": str(e),
+                "message": f"对比分析失败: {str(e)}"
+            }
+
+    async def _execute_edit(self, params: Dict[str, Any], context: Dict[str, Any]) -> Dict[str, Any]:
+        """执行文档编辑操作"""
+        try:
+            docs = context.get("source_docs", [])
+            if not docs:
+                return {
+                    "success": False,
+                    "error": "没有可用的文档",
+                    "message": "请先上传要编辑的文档"
+                }
+
+            doc = docs[0]  # 默认编辑第一个文档
+            content = doc.get("content", "")
+            original_filename = doc.get("metadata", {}).get("original_filename", "未知文档")
+
+            if not content:
+                return {
+                    "success": False,
+                    "error": "文档内容为空",
+                    "message": "该文档没有可编辑的内容"
+                }
+
+            # 使用 LLM 进行文本润色/编辑
+            prompt = f"""请对以下文档内容进行编辑处理。
+
+原文内容：
+{content[:8000]}
+
+编辑要求：
+- 润色表述，使其更加专业流畅
+- 修正明显的语法错误
+- 保持原意不变
+- 只返回编辑后的内容，不要解释
+
+请直接输出编辑后的内容："""
+
+            messages = [
+                {"role": "system", "content": "你是一个专业的文本编辑助手。请直接输出编辑后的内容。"},
+                {"role": "user", "content": prompt}
+            ]
+
+            from app.services.llm_service import llm_service
+            response = await llm_service.chat(messages=messages, temperature=0.3, max_tokens=8000)
+            edited_content = llm_service.extract_message_content(response)
+
+            return {
+                "success": True,
+                "intent": "edit",
+                "edited_content": edited_content,
+                "original_filename": original_filename,
+                "message": "文档编辑完成，内容已返回"
+            }
+
+        except Exception as e:
+            logger.error(f"编辑执行失败: {e}")
+            return {
+                "success": False,
+                "error": str(e),
+                "message": f"编辑处理失败: {str(e)}"
+            }
+
+    async def _execute_transform(self, params: Dict[str, Any], context: Dict[str, Any]) -> Dict[str, Any]:
+        """
+        执行格式转换操作
+
+        支持：
+        - Word -> Excel
+        - Excel -> Word
+        - Markdown -> Word
+        - Word -> Markdown
+        """
+        try:
+            docs = context.get("source_docs", [])
+            if not docs:
+                return {
+                    "success": False,
+                    "error": "没有可用的文档",
+                    "message": "请先上传要转换的文档"
+                }
+
+            # 获取目标格式
+            template_info = params.get("template", {})
+            target_type = template_info.get("type", "")
+
+            if not target_type:
+                # 尝试从指令中推断
+                instruction = params.get("instruction", "")
+                if "excel" in instruction.lower() or "xlsx" in instruction.lower():
+                    target_type = "xlsx"
+                elif "word" in instruction.lower() or "docx" in instruction.lower():
+                    target_type = "docx"
+                elif "markdown" in instruction.lower() or "md" in instruction.lower():
+                    target_type = "md"
+
+            if not target_type:
+                return {
+                    "success": False,
+                    "error": "未指定目标格式",
+                    "message": "请说明要转换成什么格式（如：转成Excel、转成Word）"
+                }
+
+            doc = docs[0]
+            content = doc.get("content", "")
+            structured_data = doc.get("structured_data", {})
+            original_filename = doc.get("metadata", {}).get("original_filename", "未知文档")
+
+            # 构建转换内容
+            if structured_data.get("tables"):
+                # 有表格数据，生成表格格式的内容
+                tables = structured_data.get("tables", [])
+                table_content = []
+                for i, table in enumerate(tables[:3]):  # 最多处理3个表格
+                    headers = table.get("headers", [])
+                    rows = table.get("rows", [])[:20]  # 最多20行
+                    if headers:
+                        table_content.append(f"【表格 {i+1}】")
+                        table_content.append(" | ".join(str(h) for h in headers))
+                        table_content.append(" | ".join(["---"] * len(headers)))
+                        for row in rows:
+                            if isinstance(row, list):
+                                table_content.append(" | ".join(str(c) for c in row))
+                            elif isinstance(row, dict):
+                                table_content.append(" | ".join(str(row.get(h, "")) for h in headers))
+                        table_content.append("")
+
+                if target_type == "xlsx":
+                    # 生成 Excel 格式的数据（JSON）
+                    excel_data = []
+                    for table in tables[:1]:  # 只处理第一个表格
+                        headers = table.get("headers", [])
+                        rows = table.get("rows", [])[:100]
+                        for row in rows:
+                            if isinstance(row, list):
+                                excel_data.append(dict(zip(headers, row)))
+                            elif isinstance(row, dict):
+                                excel_data.append(row)
+
+                    return {
+                        "success": True,
+                        "intent": "transform",
+                        "transform_type": "to_excel",
+                        "target_format": "xlsx",
+                        "excel_data": excel_data,
+                        "headers": headers,
+                        "message": f"已转换为 Excel 格式，包含 {len(excel_data)} 行数据"
+                    }
+                elif target_type in ["docx", "word"]:
+                    # 生成 Word 格式的文本
+                    word_content = f"# {original_filename}\n\n"
+                    word_content += "\n".join(table_content)
+
+                    return {
+                        "success": True,
+                        "intent": "transform",
+                        "transform_type": "to_word",
+                        "target_format": "docx",
+                        "content": word_content,
+                        "message": "已转换为 Word 格式"
+                    }
+                elif target_type == "md":
+                    # 生成 Markdown 格式
+                    md_content = f"# {original_filename}\n\n"
+                    md_content += "\n".join(table_content)
+
+                    return {
+                        "success": True,
+                        "intent": "transform",
+                        "transform_type": "to_markdown",
+                        "target_format": "md",
+                        "content": md_content,
+                        "message": "已转换为 Markdown 格式"
+                    }
+
+            # 无表格数据，使用纯文本内容转换
+            if target_type == "xlsx":
+                # 将文本内容转为 Excel 格式（每行作为一列）
+                lines = [line.strip() for line in content.split("\n") if line.strip()][:100]
+                excel_data = [{"行号": i+1, "内容": line} for i, line in enumerate(lines)]
+
+                return {
+                    "success": True,
+                    "intent": "transform",
+                    "transform_type": "to_excel",
+                    "target_format": "xlsx",
+                    "excel_data": excel_data,
+                    "headers": ["行号", "内容"],
+                    "message": f"已将文本内容转换为 Excel，包含 {len(excel_data)} 行"
+                }
+            elif target_type in ["docx", "word"]:
+                return {
+                    "success": True,
+                    "intent": "transform",
+                    "transform_type": "to_word",
+                    "target_format": "docx",
+                    "content": content,
+                    "message": "文档内容已准备好，可下载为 Word 格式"
+                }
+            elif target_type == "md":
+                # 简单的文本转 Markdown
+                md_lines = []
+                for line in content.split("\n"):
+                    line = line.strip()
+                    if line:
+                        # 简单处理：如果行不长且不是列表格式，作为段落
+                        if len(line) < 100 and not line.startswith(("-", "*", "1.", "2.", "3.")):
+                            md_lines.append(line)
+                        else:
+                            md_lines.append(line)
+                    else:
+                        md_lines.append("")
+
+                return {
+                    "success": True,
+                    "intent": "transform",
+                    "transform_type": "to_markdown",
+                    "target_format": "md",
+                    "content": "\n".join(md_lines),
+                    "message": "已转换为 Markdown 格式"
+                }
+
+            return {
+                "success": False,
+                "error": "不支持的目标格式",
+                "message": f"暂不支持转换为 {target_type} 格式"
+            }
+
+        except Exception as e:
+            logger.error(f"格式转换失败: {e}")
+            return {
+                "success": False,
+                "error": str(e),
+                "message": f"格式转换失败: {str(e)}"
+            }
 
 
-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")
+# 全局单例
+instruction_executor = InstructionExecutor()

backend/app/instruction/intent_parser.py

@@ -2,17 +2,51 @@
 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
+import re
+import logging
+from typing import Any, Dict, List, Optional, Tuple
+
+logger = logging.getLogger(__name__)
 
 
-class IntentParser(ABC):
-    """意图解析器抽象基类"""
+class IntentParser:
+    """意图解析器"""
+
+    # 意图类型定义
+    INTENT_EXTRACT = "extract"           # 信息提取
+    INTENT_FILL_TABLE = "fill_table"    # 填表
+    INTENT_SUMMARIZE = "summarize"       # 摘要总结
+    INTENT_QUESTION = "question"         # 问答
+    INTENT_SEARCH = "search"              # 搜索
+    INTENT_COMPARE = "compare"            # 对比分析
+    INTENT_TRANSFORM = "transform"        # 格式转换
+    INTENT_EDIT = "edit"                  # 编辑文档
+    INTENT_UNKNOWN = "unknown"            # 未知
+
+    # 意图关键词映射
+    INTENT_KEYWORDS = {
+        INTENT_EXTRACT: ["提取", "抽取", "获取", "找出", "查找", "识别", "找到"],
+        INTENT_FILL_TABLE: ["填表", "填写", "填充", "录入", "导入到表格", "填写到"],
+        INTENT_SUMMARIZE: ["总结", "摘要", "概括", "概述", "归纳", "提炼"],
+        INTENT_QUESTION: ["问答", "回答", "解释", "什么是", "为什么", "如何", "怎样", "多少", "几个"],
+        INTENT_SEARCH: ["搜索", "查找", "检索", "查询", "找"],
+        INTENT_COMPARE: ["对比", "比较", "差异", "区别", "不同"],
+        INTENT_TRANSFORM: ["转换", "转化", "变成", "转为", "导出"],
+        INTENT_EDIT: ["修改", "编辑", "调整", "改写", "润色", "优化"],
+    }
+
+    # 实体模式定义
+    ENTITY_PATTERNS = {
+        "number": [r"\d+", r"[一二三四五六七八九十百千万]+"],
+        "date": [r"\d{4}年", r"\d{1,2}月", r"\d{1,2}日"],
+        "percentage": [r"\d+(\.\d+)?%", r"\d+(\.\d+)?‰"],
+        "currency": [r"\d+(\.\d+)?万元", r"\d+(\.\d+)?亿元", r"\d+(\.\d+)?元"],
+    }
+
+    def __init__(self):
+        self.intent_history: List[Dict[str, Any]] = []
 
-    @abstractmethod
     async def parse(self, text: str) -> Tuple[str, Dict[str, Any]]:
         """
         解析自然语言指令
@@ -23,12 +57,186 @@ class IntentParser(ABC):
         Returns:
             (意图类型, 参数字典)
         """
-        pass
+        text = text.strip()
+        if not text:
+            return self.INTENT_UNKNOWN, {}
+
+        # 记录历史
+        self.intent_history.append({"text": text, "intent": None})
+
+        # 识别意图
+        intent = self._recognize_intent(text)
+
+        # 提取参数
+        params = self._extract_params(text, intent)
+
+        # 更新历史
+        if self.intent_history:
+            self.intent_history[-1]["intent"] = intent
+
+        logger.info(f"意图解析: text={text[:50]}..., intent={intent}, params={params}")
+
+        return intent, params
+
+    def _recognize_intent(self, text: str) -> str:
+        """识别意图类型"""
+        intent_scores: Dict[str, float] = {}
+
+        for intent, keywords in self.INTENT_KEYWORDS.items():
+            score = 0
+            for keyword in keywords:
+                if keyword in text:
+                    score += 1
+            if score > 0:
+                intent_scores[intent] = score
+
+        if not intent_scores:
+            return self.INTENT_UNKNOWN
+
+        # 返回得分最高的意图
+        return max(intent_scores, key=intent_scores.get)
+
+    def _extract_params(self, text: str, intent: str) -> Dict[str, Any]:
+        """提取参数"""
+        params: Dict[str, Any] = {
+            "entities": self._extract_entities(text),
+            "document_refs": self._extract_document_refs(text),
+            "field_refs": self._extract_field_refs(text),
+            "template_refs": self._extract_template_refs(text),
+        }
+
+        # 根据意图类型提取特定参数
+        if intent == self.INTENT_QUESTION:
+            params["question"] = text
+            params["focus"] = self._extract_question_focus(text)
+        elif intent == self.INTENT_FILL_TABLE:
+            params["template"] = self._extract_template_info(text)
+        elif intent == self.INTENT_EXTRACT:
+            params["target_fields"] = self._extract_target_fields(text)
+
+        return params
+
+    def _extract_entities(self, text: str) -> Dict[str, List[str]]:
+        """提取实体"""
+        entities: Dict[str, List[str]] = {}
+
+        for entity_type, patterns in self.ENTITY_PATTERNS.items():
+            matches = []
+            for pattern in patterns:
+                found = re.findall(pattern, text)
+                matches.extend(found)
+            if matches:
+                entities[entity_type] = list(set(matches))
+
+        return entities
+
+    def _extract_document_refs(self, text: str) -> List[str]:
+        """提取文档引用"""
+        # 匹配 "文档1"、"doc1"、"第一个文档" 等
+        refs = []
+
+        # 数字索引: 文档1, doc1, 第1个文档
+        num_patterns = [
+            r"[文档doc]+(\d+)",
+            r"第(\d+)个文档",
+            r"第(\d+)份",
+        ]
+        for pattern in num_patterns:
+            matches = re.findall(pattern, text.lower())
+            refs.extend([f"doc_{m}" for m in matches])
+
+        # "所有文档"、"全部文档"
+        if any(kw in text for kw in ["所有", "全部", "整个"]):
+            refs.append("all_docs")
+
+        return refs
+
+    def _extract_field_refs(self, text: str) -> List[str]:
+        """提取字段引用"""
+        fields = []
+
+        # 匹配引号内的字段名
+        quoted = re.findall(r"['\"『「]([^'\"』」]+)['\"』」]", text)
+        fields.extend(quoted)
+
+        # 匹配 "xxx字段"、"xxx列" 等
+        field_patterns = [
+            r"([^\s]+)字段",
+            r"([^\s]+)列",
+            r"([^\s]+)数据",
+        ]
+        for pattern in field_patterns:
+            matches = re.findall(pattern, text)
+            fields.extend(matches)
+
+        return list(set(fields))
+
+    def _extract_template_refs(self, text: str) -> List[str]:
+        """提取模板引用"""
+        templates = []
+
+        # 匹配 "表格模板"、"Excel模板"、"表1" 等
+        template_patterns = [
+            r"([^\s]+模板)",
+            r"表(\d+)",
+            r"([^\s]+表格)",
+        ]
+        for pattern in template_patterns:
+            matches = re.findall(pattern, text)
+            templates.extend(matches)
+
+        return list(set(templates))
+
+    def _extract_question_focus(self, text: str) -> Optional[str]:
+        """提取问题焦点"""
+        # "什么是XXX"、"XXX是什么"
+        match = re.search(r"[什么是]([^?]+)", text)
+        if match:
+            return match.group(1).strip()
+
+        # "XXX有多少"
+        match = re.search(r"([^?]+)有多少", text)
+        if match:
+            return match.group(1).strip()
+
+        return None
+
+    def _extract_template_info(self, text: str) -> Optional[Dict[str, str]]:
+        """提取模板信息"""
+        template_info: Dict[str, str] = {}
+
+        # 提取模板类型
+        if "excel" in text.lower() or "xlsx" in text.lower() or "电子表格" in text:
+            template_info["type"] = "xlsx"
+        elif "word" in text.lower() or "docx" in text.lower() or "文档" in text:
+            template_info["type"] = "docx"
+
+        return template_info if template_info else None
+
+    def _extract_target_fields(self, text: str) -> List[str]:
+        """提取目标字段"""
+        fields = []
+
+        # 匹配 "提取XXX和YYY"、"抽取XXX、YYY"
+        patterns = [
+            r"提取([^(and|,|，)+]+?)(?:和|与|、|,|plus)",
+            r"抽取([^(and|,|，)+]+?)(?:和|与|、|,|plus)",
+        ]
+
+        for pattern in patterns:
+            matches = re.findall(pattern, text)
+            fields.extend([m.strip() for m in matches if m.strip()])
+
+        return list(set(fields))
+
+    def get_intent_history(self) -> List[Dict[str, Any]]:
+        """获取意图历史"""
+        return self.intent_history
+
+    def clear_history(self):
+        """清空历史"""
+        self.intent_history = []
 
 
-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")
+# 全局单例
+intent_parser = IntentParser()