update

backend/.env

@@ -0,0 +1,56 @@
+APP_NAME=AI+合规智能中枢
+APP_VERSION=0.1.0
+DEBUG=false
+
+MILVUS_HOST=localhost
+MILVUS_PORT=19530
+MILVUS_COLLECTION=regulations
+MILVUS_DB_NAME=default
+
+MINIO_ENDPOINT=localhost:9000
+MINIO_ACCESS_KEY=minioadmin
+MINIO_SECRET_KEY=minioadmin
+MINIO_BUCKET=compliance-docs
+MINIO_SECURE=false
+
+REDIS_HOST=localhost
+REDIS_PORT=6379
+REDIS_PASSWORD=redis@123
+REDIS_DB=0
+
+POSTGRES_HOST=localhost
+POSTGRES_PORT=5432
+POSTGRES_USER=postgresql
+POSTGRES_PASSWORD=postgresql123456
+POSTGRES_DB=compliance_db
+
+EMBEDDING_MODEL=BAAI/bge-m3
+EMBEDDING_DIM=1024
+EMBEDDING_MAX_LENGTH=8192
+EMBEDDING_BATCH_SIZE=12
+EMBEDDING_USE_FP16=true
+
+CHUNK_SIZE=512
+CHUNK_OVERLAP=50
+MAX_FILE_SIZE_MB=100
+
+API_HOST=0.0.0.0
+API_PORT=8000
+
+LLM_PROVIDER=deepseek
+LLM_MODEL=deepseek-v4-flash
+LLM_MAX_TOKENS=4096
+LLM_TEMPERATURE=0.7
+
+QWEN_API_KEY=sk-fVr9KmDZNC4pGDBQj0EUWz9bDmFzNxjYC9EzZpe2bVDsxtz8
+QWEN_BASE_URL=http://6.86.80.4:30080/v1
+QWEN_MODEL=qwen3.5-plus
+QWEN_VL_MODEL=qwen3-vl-plus
+
+DEEPSEEK_API_KEY=sk-fVr9KmDZNC4pGDBQj0EUWz9bDmFzNxjYC9EzZpe2bVDsxtz8
+DEEPSEEK_BASE_URL=http://6.86.80.4:30080/v1
+DEEPSEEK_MODEL=deepseek-v4-flash
+
+RAG_TOP_K=10
+RAG_MAX_CONTEXT_TOKENS=4000
+RAG_SUMMARY_MAX_TOKENS=1024

backend/.env.example

@@ -0,0 +1,56 @@
+APP_NAME=AI+合规智能中枢
+APP_VERSION=0.1.0
+DEBUG=false
+
+MILVUS_HOST=localhost
+MILVUS_PORT=19530
+MILVUS_COLLECTION=regulations
+MILVUS_DB_NAME=default
+
+EMBEDDING_MODEL=BAAI/bge-m3
+EMBEDDING_DIM=1024
+EMBEDDING_MAX_LENGTH=8192
+EMBEDDING_BATCH_SIZE=12
+EMBEDDING_USE_FP16=true
+
+MINIO_ENDPOINT=localhost:9000
+MINIO_ACCESS_KEY=minioadmin
+MINIO_SECRET_KEY=minioadmin123
+MINIO_BUCKET=compliance-docs
+MINIO_SECURE=false
+
+REDIS_HOST=localhost
+REDIS_PORT=6379
+REDIS_PASSWORD=
+REDIS_DB=0
+
+POSTGRES_HOST=localhost
+POSTGRES_PORT=5432
+POSTGRES_USER=compliance
+POSTGRES_PASSWORD=compliance123
+POSTGRES_DB=compliance_db
+
+CHUNK_SIZE=512
+CHUNK_OVERLAP=50
+MAX_FILE_SIZE_MB=100
+
+API_HOST=0.0.0.0
+API_PORT=8000
+
+LLM_PROVIDER=deepseek
+LLM_MODEL=deepseek-v4-flash
+LLM_MAX_TOKENS=4096
+LLM_TEMPERATURE=0.7
+
+QWEN_API_KEY=your_api_key_here
+DEEPSEEK_API_KEY=your_api_key_here
+QWEN_BASE_URL=http://6.86.80.4:30080/v1
+DEEPSEEK_BASE_URL=http://6.86.80.4:30080/v1
+
+QWEN_MODEL=qwen3.5-plus
+QWEN_VL_MODEL=qwen3-vl-plus
+DEEPSEEK_MODEL=deepseek-v4-flash
+
+RAG_TOP_K=10
+RAG_MAX_CONTEXT_TOKENS=4000
+RAG_SUMMARY_MAX_TOKENS=1024

backend/.gitignore

@@ -0,0 +1,10 @@
+# Python-generated files
+__pycache__/
+*.py[oc]
+build/
+dist/
+wheels/
+*.egg-info
+
+# Virtual environments
+.venv

backend/.python-version

@@ -0,0 +1 @@
+3.9

backend/Dockerfile

@@ -0,0 +1,18 @@
+FROM python:3.11-slim
+
+WORKDIR /app
+
+# 安装依赖
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# 复制代码
+COPY app/ ./app/
+COPY data/ ./data/
+
+# 环境变量
+ENV API_HOST=0.0.0.0
+ENV API_PORT=8000
+
+# 启动命令
+CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "8000"]

backend/README.md

@@ -0,0 +1,50 @@
+# AI+合规智能中枢后端
+
+`backend` 已承接原 `src` 的完整 FastAPI 后端能力，当前正式入口为 `app.main:app`。
+
+## 启动
+
+```bash
+pip install -r backend/requirements.txt
+PYTHONPATH=backend uvicorn app.main:app --host 0.0.0.0 --port 8000 --reload
+```
+
+也可以直接使用根目录脚本：
+
+```bash
+./start_api.sh
+```
+
+## 主要接口
+
+- `GET /health`
+- `GET /`
+- `POST /api/v1/documents/upload`
+- `GET /api/v1/documents/list`
+- `GET /api/v1/documents/management-list`
+- `GET /api/v1/documents/download/{doc_id}`
+- `POST /api/v1/knowledge/search`
+- `POST /api/v1/knowledge/retrieval`
+- `POST /api/v1/agent/ask`
+- `POST /api/v1/agent/chat`
+- `GET /api/v1/agent/chat/stream`
+
+## 目录说明
+
+```text
+backend/
+├── app/
+│   ├── api/          # FastAPI 路由与模型
+│   ├── config/       # 配置与日志
+│   ├── services/     # 文档处理、LLM、RAG、存储
+│   └── workers/      # 任务相关代码
+├── .env.example
+├── requirements.txt
+└── main.py
+```
+
+## 说明
+
+- `backend/app/api/main.py` 来自原 `src/api/main.py`，已切换为 `app.*` 导入。
+- 路由前缀保持为 `/api/v1`，以兼容当前前端。
+- 原 `backend/app/api/routes/docs.py`、`rag.py`、`compliance.py`、`status.py` 仍保留在仓库中，但不再作为主路由入口。

backend/app/__init__.py

@@ -0,0 +1,3 @@
+from .main import app
+
+__all__ = ["app"]

backend/app/api/__init__.py

@@ -0,0 +1,2 @@
+# src/api/__init__.py
+"""API接口模块"""

backend/app/api/main.py

@@ -0,0 +1,99 @@
+"""FastAPI application entrypoint."""
+
+from contextlib import asynccontextmanager
+
+from fastapi import FastAPI, Request
+from fastapi.middleware.cors import CORSMiddleware
+from fastapi.responses import JSONResponse
+from loguru import logger
+
+from app.api.models import ErrorResponse
+from app.api.routes import api_router
+from app.config.logging import setup_logging
+from app.config.settings import settings
+from app.services.llm.llm_factory import LLMFactory
+
+setup_logging(level="INFO" if not settings.debug else "DEBUG")
+
+
+@asynccontextmanager
+async def lifespan(app: FastAPI):
+    """Application lifecycle hooks."""
+    logger.info(f"启动 {settings.app_name} v{settings.app_version}")
+    logger.info(f"调试模式: {settings.debug}")
+    logger.info("预加载LLM客户端...")
+    LLMFactory.preload_clients(["qwen", "deepseek"])
+
+    yield
+
+    logger.info("应用关闭，执行清理...")
+    LLMFactory.cleanup()
+
+
+app = FastAPI(
+    title=settings.app_name,
+    description=(
+        "AI+合规智能中枢 - 法律法规文档解析入库功能\n\n"
+        "支持PDF/DOCX文档解析、智能分块、向量嵌入、Milvus存储"
+    ),
+    version=settings.app_version,
+    lifespan=lifespan,
+    docs_url="/docs",
+    redoc_url="/redoc",
+)
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+app.include_router(api_router, prefix="/api/v1")
+
+
+@app.exception_handler(Exception)
+async def global_exception_handler(request: Request, exc: Exception):
+    """Global exception handler."""
+    logger.error(f"未处理的异常: {exc}")
+    return JSONResponse(
+        status_code=500,
+        content=ErrorResponse(
+            error="InternalServerError",
+            message=str(exc),
+        ).model_dump(),
+    )
+
+
+@app.get("/health", tags=["health"])
+async def health_check():
+    """Health check endpoint."""
+    return {
+        "status": "healthy",
+        "app": settings.app_name,
+        "version": settings.app_version,
+    }
+
+
+@app.get("/", tags=["root"])
+async def root():
+    """Root endpoint."""
+    return {
+        "message": f"Welcome to {settings.app_name}",
+        "version": settings.app_version,
+        "docs": "/docs",
+        "health": "/health",
+    }
+
+
+if __name__ == "__main__":
+    import uvicorn
+
+    uvicorn.run(
+        "app.api.main:app",
+        host=settings.api_host,
+        port=settings.api_port,
+        reload=settings.debug,
+        log_level="info",
+    )

backend/app/api/models/__init__.py

@@ -0,0 +1,22 @@
+# src/api/models/__init__.py
+"""API数据模型"""
+
+from .document import (
+    DocumentUploadRequest,
+    DocumentUploadResponse,
+    SearchRequest,
+    SearchResultItem,
+    SearchResponse,
+    DocumentStatusResponse,
+    ErrorResponse
+)
+
+__all__ = [
+    "DocumentUploadRequest",
+    "DocumentUploadResponse",
+    "SearchRequest",
+    "SearchResultItem",
+    "SearchResponse",
+    "DocumentStatusResponse",
+    "ErrorResponse"
+]

backend/app/api/models/document.py

@@ -0,0 +1,63 @@
+# src/api/models/document.py
+"""文档相关Pydantic数据模型"""
+
+from pydantic import BaseModel, Field
+from typing import List, Optional, Dict, Any
+from datetime import datetime
+
+
+class DocumentUploadRequest(BaseModel):
+    """文档上传请求"""
+    doc_name: Optional[str] = Field(None, description="文档名称")
+    regulation_type: Optional[str] = Field(None, description="法规类型")
+    version: Optional[str] = Field(None, description="文档版本")
+
+
+class DocumentUploadResponse(BaseModel):
+    """文档上传响应"""
+    doc_id: str = Field(..., description="文档ID")
+    doc_name: str = Field(..., description="文档名称")
+    status: str = Field(..., description="处理状态")
+    message: str = Field(default="", description="状态消息")
+    num_chunks: int = Field(default=0, description="分块数量")
+    summary: str = Field(default="", description="LLM生成的文档摘要")
+    summary_latency_ms: int = Field(default=0, description="摘要生成耗时(ms)")
+    timestamp: datetime = Field(default_factory=datetime.now, description="时间戳")
+
+
+class SearchRequest(BaseModel):
+    """检索请求"""
+    query: str = Field(..., description="查询文本")
+    top_k: int = Field(default=10, description="返回结果数量")
+    filters: Optional[str] = Field(None, description="过滤条件")
+
+
+class SearchResultItem(BaseModel):
+    """单个检索结果"""
+    id: int = Field(..., description="记录ID")
+    content: str = Field(..., description="内容")
+    score: float = Field(..., description="相似度分数")
+    metadata: Dict[str, Any] = Field(default_factory=dict, description="元数据")
+
+
+class SearchResponse(BaseModel):
+    """检索响应"""
+    query: str = Field(..., description="查询文本")
+    total: int = Field(..., description="结果总数")
+    results: List[SearchResultItem] = Field(default_factory=list, description="结果列表")
+    timestamp: datetime = Field(default_factory=datetime.now, description="时间戳")
+
+
+class DocumentStatusResponse(BaseModel):
+    """文档状态响应"""
+    doc_id: str = Field(..., description="文档ID")
+    status: str = Field(..., description="状态")
+    num_chunks: Optional[int] = Field(None, description="分块数量")
+    timestamp: datetime = Field(default_factory=datetime.now, description="时间戳")
+
+
+class ErrorResponse(BaseModel):
+    """错误响应"""
+    error: str = Field(..., description="错误类型")
+    message: str = Field(..., description="错误消息")
+    timestamp: datetime = Field(default_factory=datetime.now, description="时间戳")

backend/app/api/routes/__init__.py

@@ -0,0 +1,17 @@
+# src/api/routes/__init__.py
+"""API路由模块"""
+
+from fastapi import APIRouter
+from .documents import router as documents_router
+from .knowledge import router as knowledge_router
+from .agent import router as agent_router
+
+# 主路由
+api_router = APIRouter()
+
+# 注册子路由
+api_router.include_router(documents_router)
+api_router.include_router(knowledge_router)
+api_router.include_router(agent_router)
+
+__all__ = ["api_router", "documents_router", "knowledge_router", "agent_router"]

backend/app/api/routes/agent.py

@@ -0,0 +1,449 @@
+# src/api/routes/agent.py
+"""Agent API接口 - 问答对话接口"""
+
+from fastapi import APIRouter, HTTPException, Depends
+from fastapi.responses import StreamingResponse
+from pydantic import BaseModel, Field
+from typing import List, Dict, Optional, AsyncGenerator
+from loguru import logger
+import json
+import asyncio
+
+from app.services.agent.qa_agent import QAAgent, AgentConfig
+from app.services.agent.session_manager import SessionManager
+from app.config.settings import settings
+
+
+router = APIRouter(prefix="/agent", tags=["agent"])
+
+# 会话管理器（全局实例）
+session_manager = SessionManager()
+
+
+# ===== Pydantic Models =====
+
+class AskRequest(BaseModel):
+    """单次问答请求"""
+    query: str = Field(..., description="用户问题", min_length=1, max_length=2000)
+    filters: Optional[str] = Field(None, description="检索过滤条件")
+    provider: Optional[str] = Field(None, description="LLM提供商 (qwen/deepseek)")
+    model: Optional[str] = Field(None, description="LLM模型名称")
+    top_k: Optional[int] = Field(None, description="检索数量", ge=1, le=20)
+    prompt_template: Optional[str] = Field(None, description="Prompt模板名称")
+
+
+class AskResponse(BaseModel):
+    """问答响应"""
+    answer: str
+    sources: List[Dict] = []
+    model: str = ""
+    latency_ms: int = 0
+    retrieved_count: int = 0
+    context_tokens: int = 0
+    truncated: bool = False
+    error: Optional[str] = None
+
+
+class ChatRequest(BaseModel):
+    """多轮对话请求"""
+    query: str = Field(..., description="用户问题", min_length=1, max_length=2000)
+    session_id: Optional[str] = Field(None, description="会话ID（首次对话可不传）")
+    filters: Optional[str] = Field(None, description="检索过滤条件")
+    provider: Optional[str] = Field(None, description="LLM提供商")
+    model: Optional[str] = Field(None, description="LLM模型名称")
+
+
+class ChatResponse(BaseModel):
+    """多轮对话响应"""
+    session_id: str
+    answer: str
+    sources: List[Dict] = []
+    model: str = ""
+    latency_ms: int = 0
+    message_count: int = 0
+
+
+class SessionInfo(BaseModel):
+    """会话信息"""
+    session_id: str
+    message_count: int
+    created_at: int
+    updated_at: int
+
+
+class FeedbackRequest(BaseModel):
+    """反馈请求"""
+    session_id: str
+    message_index: int
+    rating: int = Field(..., ge=1, le=5, description="评分 1-5")
+    comment: Optional[str] = Field(None, description="反馈内容")
+
+
+class TemplateListResponse(BaseModel):
+    """模板列表响应"""
+    templates: Dict[str, str]
+
+
+# ===== API Endpoints =====
+
+@router.post("/ask", response_model=AskResponse)
+async def ask_question(request: AskRequest):
+    """
+    单次问答接口
+
+    不保存会话历史，适合单次查询场景。
+    """
+    logger.info(f"收到问答请求: {request.query}")
+
+    try:
+        # 构建Agent配置
+        config = AgentConfig(
+            llm_provider=request.provider or settings.llm_provider,
+            llm_model=request.model or settings.llm_model,
+            top_k=request.top_k or settings.rag_top_k
+        )
+
+        # 创建Agent并执行问答
+        agent = QAAgent(config)
+        response = agent.ask(
+            query=request.query,
+            filters=request.filters,
+            prompt_template=request.prompt_template
+        )
+        agent.close()
+
+        return AskResponse(
+            answer=response.answer,
+            sources=response.sources,
+            model=response.model,
+            latency_ms=response.latency_ms,
+            retrieved_count=response.retrieved_count,
+            context_tokens=response.context_tokens,
+            truncated=response.truncated,
+            error=response.error
+        )
+
+    except Exception as e:
+        logger.error(f"问答失败: {e}")
+        raise HTTPException(status_code=500, detail=str(e))
+
+
+@router.post("/chat", response_model=ChatResponse)
+async def chat_with_session(request: ChatRequest):
+    """
+    多轮对话接口
+
+    支持会话历史记录，适合连续对话场景。
+    """
+    logger.info(f"收到对话请求: session={request.session_id}, query={request.query}")
+
+    try:
+        # 获取或创建会话
+        if request.session_id:
+            session = session_manager.get_session(request.session_id)
+            if not session:
+                raise HTTPException(status_code=404, detail="会话不存在或已过期")
+        else:
+            session = session_manager.create_session()
+
+        # 添加用户消息
+        session.add_user_message(request.query)
+
+        # 执行问答
+        config = AgentConfig(
+            llm_provider=request.provider or settings.llm_provider,
+            llm_model=request.model or settings.llm_model
+        )
+
+        agent = QAAgent(config)
+        response = agent.ask(
+            query=request.query,
+            filters=request.filters
+        )
+        agent.close()
+
+        # 添加助手消息
+        session.add_assistant_message(
+            response.answer,
+            response.sources
+        )
+
+        return ChatResponse(
+            session_id=session.session_id,
+            answer=response.answer,
+            sources=response.sources,
+            model=response.model,
+            latency_ms=response.latency_ms,
+            message_count=session.message_count
+        )
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"对话失败: {e}")
+        raise HTTPException(status_code=500, detail=str(e))
+
+
+@router.get("/chat/stream")
+async def chat_stream_get(
+    query: str,
+    session_id: Optional[str] = None,
+    filters: Optional[str] = None,
+    provider: Optional[str] = None,
+    model: Optional[str] = None
+):
+    """
+    流式对话接口（SSE）- GET版本
+
+    EventSource只能发送GET请求，因此提供此接口。
+    query参数通过URL传递。
+
+    SSE事件格式：
+    - event: session   - 会话ID
+    - event: status    - 状态更新（检索中、生成中）
+    - event: sources   - 引用来源
+    - event: content   - 回答内容片段
+    - event: done      - 完成，包含统计信息
+    - event: error     - 错误信息
+    """
+    logger.info(f"收到GET流式对话请求: session={session_id}, query={query}")
+
+    async def generate_sse() -> AsyncGenerator[str, None]:
+        """生成SSE事件流"""
+        try:
+            # 获取或创建会话
+            if session_id:
+                session = session_manager.get_session(session_id)
+                if not session:
+                    yield f"event: error\ndata: 会话不存在或已过期\n\n"
+                    return
+            else:
+                session = session_manager.create_session()
+
+            # 发送session_id
+            yield f"event: session\ndata: {json.dumps({'session_id': session.session_id})}\n\n"
+
+            # 添加用户消息
+            session.add_user_message(query)
+
+            # 创建Agent
+            config = AgentConfig(
+                llm_provider=provider or settings.llm_provider,
+                llm_model=model or settings.llm_model
+            )
+
+            agent = QAAgent(config)
+
+            # 执行流式问答
+            full_answer = ""
+            sources = []
+            done_data = {}
+
+            for event_data in agent.ask_stream(
+                query=query,
+                filters=filters
+            ):
+                event_type = event_data.get("event", "content")
+                data = event_data.get("data", "")
+
+                # 收集完整回答和来源
+                if event_type == "content":
+                    full_answer += str(data)
+                elif event_type == "sources":
+                    sources = data
+                elif event_type == "done":
+                    done_data = data
+
+                # 发送SSE事件
+                if isinstance(data, (dict, list)):
+                    yield f"event: {event_type}\ndata: {json.dumps(data)}\n\n"
+                else:
+                    yield f"event: {event_type}\ndata: {data}\n\n"
+
+                # 小延迟让其他任务有机会执行
+                await asyncio.sleep(0)
+
+            agent.close()
+
+            # 保存到会话历史
+            session.add_assistant_message(full_answer, sources)
+
+        except Exception as e:
+            logger.error(f"流式对话失败: {e}")
+            yield f"event: error\ndata: {str(e)}\n\n"
+
+    return StreamingResponse(
+        generate_sse(),
+        media_type="text/event-stream",
+        headers={
+            "Cache-Control": "no-cache",
+            "Connection": "keep-alive",
+            "X-Accel-Buffering": "no"  # 禁用nginx缓冲
+        }
+    )
+
+
+@router.post("/chat/stream")
+async def chat_stream(request: ChatRequest):
+    """
+    流式对话接口（SSE）
+
+    返回Server-Sent Events格式的流式响应，用户可实时看到思考过程和回答生成。
+
+    SSE事件格式：
+    - event: status    - 状态更新（检索中、生成中）
+    - event: sources   - 引用来源
+    - event: content   - 回答内容片段
+    - event: done      - 完成，包含统计信息
+    - event: error     - 错误信息
+    """
+    logger.info(f"收到流式对话请求: session={request.session_id}, query={request.query}")
+
+    async def generate_sse() -> AsyncGenerator[str, None]:
+        """生成SSE事件流"""
+        try:
+            # 获取或创建会话
+            if request.session_id:
+                session = session_manager.get_session(request.session_id)
+                if not session:
+                    yield f"event: error\ndata: 会话不存在或已过期\n\n"
+                    return
+            else:
+                session = session_manager.create_session()
+
+            # 发送session_id
+            yield f"event: session\ndata: {json.dumps({'session_id': session.session_id})}\n\n"
+
+            # 添加用户消息
+            session.add_user_message(request.query)
+
+            # 创建Agent
+            config = AgentConfig(
+                llm_provider=request.provider or settings.llm_provider,
+                llm_model=request.model or settings.llm_model
+            )
+
+            agent = QAAgent(config)
+
+            # 执行流式问答
+            full_answer = ""
+            sources = []
+            done_data = {}
+
+            for event_data in agent.ask_stream(
+                query=request.query,
+                filters=request.filters
+            ):
+                event_type = event_data.get("event", "content")
+                data = event_data.get("data", "")
+
+                # 收集完整回答和来源
+                if event_type == "content":
+                    full_answer += str(data)
+                elif event_type == "sources":
+                    sources = data
+                elif event_type == "done":
+                    done_data = data
+
+                # 发送SSE事件
+                if isinstance(data, (dict, list)):
+                    yield f"event: {event_type}\ndata: {json.dumps(data)}\n\n"
+                else:
+                    yield f"event: {event_type}\ndata: {data}\n\n"
+
+                # 小延迟让其他任务有机会执行
+                await asyncio.sleep(0)
+
+            agent.close()
+
+            # 保存到会话历史
+            session.add_assistant_message(full_answer, sources)
+
+        except Exception as e:
+            logger.error(f"流式对话失败: {e}")
+            yield f"event: error\ndata: {str(e)}\n\n"
+
+    return StreamingResponse(
+        generate_sse(),
+        media_type="text/event-stream",
+        headers={
+            "Cache-Control": "no-cache",
+            "Connection": "keep-alive",
+            "X-Accel-Buffering": "no"  # 禁用nginx缓冲
+        }
+    )
+
+
+@router.get("/session/{session_id}", response_model=SessionInfo)
+async def get_session_info(session_id: str):
+    """获取会话信息"""
+    session = session_manager.get_session(session_id)
+    if not session:
+        raise HTTPException(status_code=404, detail="会话不存在或已过期")
+
+    return SessionInfo(
+        session_id=session.session_id,
+        message_count=session.message_count,
+        created_at=session.created_at,
+        updated_at=session.updated_at
+    )
+
+
+@router.get("/session/{session_id}/history")
+async def get_session_history(session_id: str, max_turns: int = 5):
+    """获取会话历史"""
+    session = session_manager.get_session(session_id)
+    if not session:
+        raise HTTPException(status_code=404, detail="会话不存在或已过期")
+
+    history = session.get_history(max_turns)
+    return {"session_id": session_id, "history": history}
+
+
+@router.delete("/session/{session_id}")
+async def delete_session(session_id: str):
+    """删除会话"""
+    success = session_manager.delete_session(session_id)
+    if not success:
+        raise HTTPException(status_code=404, detail="会话不存在")
+
+    return {"message": "会话已删除", "session_id": session_id}
+
+
+@router.get("/sessions", response_model=List[SessionInfo])
+async def list_sessions():
+    """列出所有活跃会话"""
+    sessions = session_manager.list_sessions()
+    return [SessionInfo(**s) for s in sessions]
+
+
+@router.post("/feedback")
+async def submit_feedback(request: FeedbackRequest):
+    """提交问答反馈"""
+    session = session_manager.get_session(request.session_id)
+    if not session:
+        raise HTTPException(status_code=404, detail="会话不存在")
+
+    # 记录反馈（实际应用中可存储到数据库）
+    logger.info(f"收到反馈: session={request.session_id}, rating={request.rating}, comment={request.comment}")
+
+    return {"message": "反馈已记录", "rating": request.rating}
+
+
+@router.get("/templates", response_model=TemplateListResponse)
+async def list_prompt_templates():
+    """列出可用的Prompt模板"""
+    from app.services.rag.prompt_templates import PromptTemplates
+
+    templates = PromptTemplates.list_templates()
+    return TemplateListResponse(templates=templates)
+
+
+@router.get("/models")
+async def list_available_models():
+    """列出可用的LLM模型"""
+    from app.services.llm import LLMFactory
+
+    factory = LLMFactory()
+    models = factory.list_available_providers()
+    return {"models": models}

backend/app/api/routes/compliance.py

@@ -0,0 +1,96 @@
+from fastapi import APIRouter, UploadFile, File, HTTPException
+from sse_starlette.sse import EventSourceResponse
+import uuid
+import os
+import json
+import asyncio
+from app.schemas.compliance import (
+    AnalyzeResponse,
+    ComplianceChatRequest,
+)
+from app.services.mock_data import (
+    generate_task_id,
+    get_mock_compliance_result,
+    get_mock_compliance_chat_response,
+)
+
+router = APIRouter(prefix="/compliance", tags=["合规分析"])
+
+# 临时存储分析任务
+tasks_store: dict[str, dict] = {}
+
+
+@router.post("/analyze", response_model=AnalyzeResponse)
+async def analyze_document(file: UploadFile = File(...)):
+    """上传设计方案进行分析"""
+    # 生成任务ID
+    task_id = generate_task_id()
+
+    # 保存文件
+    raw_dir = "/airegulation/demo-mao/backend/data/raw"
+    os.makedirs(raw_dir, exist_ok=True)
+    file_path = os.path.join(raw_dir, f"compliance_{task_id}_{file.filename}")
+
+    content = await file.read()
+    with open(file_path, "wb") as f:
+        f.write(content)
+
+    # 记录任务
+    tasks_store[task_id] = {
+        "task_id": task_id,
+        "file_path": file_path,
+        "status": "processing",
+        "result": None,
+    }
+
+    # 模拟异步处理完成（立即返回结果）
+    # 实际应用中这应该是后台任务
+    tasks_store[task_id]["status"] = "completed"
+    tasks_store[task_id]["result"] = get_mock_compliance_result(task_id)
+
+    return AnalyzeResponse(task_id=task_id)
+
+
+@router.get("/result/{task_id}")
+async def get_result(task_id: str):
+    """获取分析结果"""
+    if task_id not in tasks_store:
+        # 如果任务ID不存在，返回默认mock结果
+        return get_mock_compliance_result(task_id)
+
+    task = tasks_store[task_id]
+
+    if task["status"] == "processing":
+        return {"status": "processing", "message": "分析进行中"}
+
+    return task["result"]
+
+
+@router.post("/chat/{segment_id}")
+async def compliance_chat(segment_id: int, request: ComplianceChatRequest):
+    """针对段落进行合规对话"""
+    # 根据segment_id获取对应的intent
+    intent_map = {
+        1: "车身结构设计",
+        2: "动力系统配置",
+        3: "安全配置设计",
+    }
+    intent = intent_map.get(segment_id, "车身结构设计")
+
+    async def generate():
+        # 获取预设响应
+        response = get_mock_compliance_chat_response(intent, request.query)
+
+        # 流式输出响应
+        sentences = response.split("\n\n")
+        for sentence in sentences:
+            if sentence.strip():
+                chunks = sentence.split("\n")
+                for chunk in chunks:
+                    if chunk.strip():
+                        await asyncio.sleep(0.05)
+                        yield {"event": "message", "data": json.dumps({"type": "chunk", "text": chunk + "\n"})}
+
+        yield {"event": "message", "data": json.dumps({"type": "done"})}
+
+    return EventSourceResponse(generate())

backend/app/api/routes/docs.py

@@ -0,0 +1,115 @@
+from fastapi import APIRouter, UploadFile, File, HTTPException
+import os
+import uuid
+from datetime import datetime
+from app.schemas.doc import (
+    DocumentUploadResponse,
+    DocumentListResponse,
+    DocumentInfo,
+    ParseResponse,
+    EmbedResponse,
+)
+from app.services.mock_data import get_mock_documents, generate_doc_id
+
+router = APIRouter(prefix="/docs", tags=["文档管理"])
+
+# 临时存储文档信息（包含预设的mock文档）
+documents_store: dict[str, dict] = {}
+
+# 初始化时加载mock文档
+for doc in get_mock_documents():
+    documents_store[doc["id"]] = doc
+
+
+@router.post("/upload", response_model=DocumentUploadResponse)
+async def upload_document(file: UploadFile = File(...)):
+    """上传法规文档"""
+    # 检查文件格式
+    allowed_ext = [".pdf", ".docx", ".doc", ".txt"]
+    ext = os.path.splitext(file.filename)[1].lower()
+    if ext not in allowed_ext:
+        raise HTTPException(400, f"Unsupported file format: {ext}")
+
+    # 生成文档ID
+    doc_id = generate_doc_id()
+
+    # 保存文件
+    raw_dir = "/airegulation/demo-mao/backend/data/raw"
+    os.makedirs(raw_dir, exist_ok=True)
+    file_path = os.path.join(raw_dir, f"{doc_id}_{file.filename}")
+
+    content = await file.read()
+    with open(file_path, "wb") as f:
+        f.write(content)
+
+    # 记录文档信息
+    documents_store[doc_id] = {
+        "id": doc_id,
+        "name": file.filename,
+        "path": file_path,
+        "size": len(content),
+        "status": "uploaded",
+        "chunks": 0,
+        "created_at": datetime.now(),
+    }
+
+    return DocumentUploadResponse(
+        doc_id=doc_id,
+        filename=file.filename,
+        size=len(content),
+    )
+
+
+@router.get("/list", response_model=DocumentListResponse)
+async def list_documents():
+    """获取已索引文档列表"""
+    docs = [
+        DocumentInfo(
+            id=d["id"],
+            name=d["name"],
+            chunks=d["chunks"],
+            status=d["status"],
+            created_at=d.get("created_at"),
+        )
+        for d in documents_store.values()
+    ]
+    return DocumentListResponse(docs=docs)
+
+
+@router.post("/parse/{doc_id}", response_model=ParseResponse)
+async def parse_document(doc_id: str):
+    """解析文档并分块"""
+    if doc_id not in documents_store:
+        raise HTTPException(404, "Document not found")
+
+    doc = documents_store[doc_id]
+    # 模拟解析逻辑
+    doc["status"] = "parsed"
+    # 根据文件大小计算chunks数量
+    file_size = doc.get("size", 100000)
+    doc["chunks"] = max(20, file_size // 8000)
+
+    return ParseResponse(doc_id=doc_id, chunks=doc["chunks"])
+
+
+@router.post("/embed/{doc_id}", response_model=EmbedResponse)
+async def embed_document(doc_id: str):
+    """嵌入并存入向量库"""
+    if doc_id not in documents_store:
+        raise HTTPException(404, "Document not found")
+
+    doc = documents_store[doc_id]
+    # 模拟嵌入逻辑
+    doc["status"] = "indexed"
+
+    return EmbedResponse(doc_id=doc_id, vectors=doc["chunks"])
+
+
+@router.delete("/delete/{doc_id}")
+async def delete_document(doc_id: str):
+    """删除文档"""
+    if doc_id not in documents_store:
+        raise HTTPException(404, "Document not found")
+
+    del documents_store[doc_id]
+    return {"success": True}

backend/app/api/routes/documents.py

@@ -0,0 +1,291 @@
+# src/api/routes/documents.py
+"""文档上传与处理接口"""
+
+from fastapi import APIRouter, UploadFile, File, Form, HTTPException
+from fastapi.responses import FileResponse, StreamingResponse
+from typing import Optional
+import os
+import uuid
+import tempfile
+from pathlib import Path
+from loguru import logger
+from io import BytesIO
+from urllib.parse import quote
+
+from ..models import DocumentUploadResponse, ErrorResponse
+from app.services.document_processor import DocumentProcessor
+from app.services.storage.minio_client import MinIOClient
+from app.config.settings import settings
+
+router = APIRouter(prefix="/documents", tags=["documents"])
+
+# MinIO客户端（用于文档存储）
+minio_client: Optional[MinIOClient] = None
+
+
+def get_minio_client() -> MinIOClient:
+    """获取MinIO客户端实例"""
+    global minio_client
+    if minio_client is None:
+        minio_client = MinIOClient()
+        minio_client.connect()
+        minio_client.ensure_bucket()
+    return minio_client
+
+
+def _build_document_records(limit: Optional[int] = None):
+    """构建文档列表记录，支持按最近更新时间倒序截断。"""
+    minio = get_minio_client()
+
+    document_records = []
+    objects = minio.client.list_objects(minio.bucket, recursive=True)
+    for obj in objects:
+        parts = obj.object_name.split("/", 1)
+        if len(parts) != 2:
+            continue
+
+        doc_id, filename = parts
+        last_modified = getattr(obj, "last_modified", None)
+        document_records.append({
+            "doc_id": doc_id,
+            "filename": filename,
+            "size": getattr(obj, "size", 0) or 0,
+            "object_name": obj.object_name,
+            "download_url": f"/api/v1/documents/download/{doc_id}",
+            "last_modified": last_modified.isoformat() if last_modified else None,
+            "_sort_key": last_modified.timestamp() if last_modified else 0,
+        })
+
+    document_records.sort(key=lambda item: item["_sort_key"], reverse=True)
+    if limit is not None:
+        document_records = document_records[:limit]
+
+    for item in document_records:
+        item.pop("_sort_key", None)
+
+    return document_records
+
+
+@router.post("/upload", response_model=DocumentUploadResponse)
+async def upload_document(
+    file: UploadFile = File(..., description="上传的文档文件"),
+    doc_name: Optional[str] = Form(None, description="文档名称"),
+    regulation_type: Optional[str] = Form(None, description="法规类型"),
+    version: Optional[str] = Form(None, description="文档版本"),
+    generate_summary: bool = Form(False, description="是否生成摘要（默认不生成，可节省约60秒）")
+):
+    """
+    上传文档并处理
+
+    支持格式：PDF、DOCX、DOC
+    处理流程：解析 → 分块 → 嵌入 → 入库（摘要可选）
+    文件存储：MinIO对象存储
+
+    参数说明：
+    - generate_summary: 是否生成LLM摘要，默认False。勾选后处理时间增加约60秒。
+    """
+    # 验证文件类型
+    ext = os.path.splitext(file.filename)[1].lower()
+    if ext not in [".pdf", ".docx", ".doc"]:
+        raise HTTPException(
+            status_code=400,
+            detail=f"不支持的文件类型: {ext}，仅支持PDF、DOCX、DOC"
+        )
+
+    # 验证文件大小
+    if file.size and file.size > settings.max_file_size_mb * 1024 * 1024:
+        raise HTTPException(
+            status_code=400,
+            detail=f"文件过大，最大支持{settings.max_file_size_mb}MB"
+        )
+
+    # 生成文档ID
+    doc_id = str(uuid.uuid4())[:8]
+
+    # 文档名称
+    final_doc_name = doc_name or file.filename
+
+    # MinIO对象名称
+    object_name = f"{doc_id}/{file.filename}"
+
+    logger.info(f"接收到文件上传: {final_doc_name}, 类型: {ext}, doc_id={doc_id}")
+
+    try:
+        # 读取文件内容
+        content = await file.read()
+
+        # 保存临时文件用于处理
+        temp_dir = tempfile.gettempdir()
+        temp_path = os.path.join(temp_dir, f"{doc_id}_{file.filename}")
+
+        with open(temp_path, "wb") as f:
+            f.write(content)
+
+        logger.info(f"临时文件已保存到: {temp_path}")
+
+        # 上传到MinIO
+        minio = get_minio_client()
+        upload_success = minio.upload_bytes(
+            data=content,
+            object_name=object_name,
+            content_type=minio._get_content_type(file.filename),
+            metadata={
+                "doc_id": doc_id  # 仅传递ASCII安全的metadata
+            }
+        )
+
+        if upload_success:
+            logger.success(f"文件已上传到MinIO: {object_name}")
+        else:
+            logger.warning(f"MinIO上传失败，仅使用本地临时文件")
+
+        # 处理文档（传入相同的doc_id，保持一致性）
+        processor = DocumentProcessor(generate_summary=generate_summary)
+        result = processor.process(
+            file_path=temp_path,
+            doc_id=doc_id,  # 使用相同的doc_id
+            doc_name=final_doc_name,
+            regulation_type=regulation_type or "",
+            version=version or ""
+        )
+        processor.close()
+
+        # 清理临时文件
+        try:
+            os.remove(temp_path)
+        except:
+            pass
+
+        if result.success:
+            return DocumentUploadResponse(
+                doc_id=result.doc_id,
+                doc_name=result.doc_name,
+                status="success",
+                message=result.message,
+                num_chunks=result.num_chunks,
+                summary=result.summary,
+                summary_latency_ms=result.summary_latency_ms
+            )
+        else:
+            raise HTTPException(
+                status_code=500,
+                detail=result.message
+            )
+
+    except Exception as e:
+        logger.error(f"文档处理失败: {e}")
+        raise HTTPException(
+            status_code=500,
+            detail=f"文档处理失败: {str(e)}"
+        )
+
+
+@router.get("/status/{doc_id}", response_model=DocumentUploadResponse)
+async def get_document_status(doc_id: str):
+    """
+    查询文档处理状态
+
+    Args:
+        doc_id: 文档ID
+    """
+    # TODO: 实现状态查询（需要数据库支持）
+    return DocumentUploadResponse(
+        doc_id=doc_id,
+        doc_name="",
+        status="unknown",
+        message="状态查询功能待实现"
+    )
+
+
+@router.get("/download/{doc_id}")
+async def download_document(doc_id: str):
+    """
+    下载文档（从MinIO获取）
+
+    Args:
+        doc_id: 文档ID
+
+    Returns:
+        文件下载响应
+    """
+    logger.info(f"请求下载文档: doc_id={doc_id}")
+
+    try:
+        minio = get_minio_client()
+
+        # 查找该doc_id下的文件（MinIO对象名称格式: {doc_id}/{filename}）
+        objects = minio.list_objects(prefix=f"{doc_id}/")
+
+        if not objects:
+            logger.warning(f"MinIO中未找到文档: doc_id={doc_id}")
+            raise HTTPException(
+                status_code=404,
+                detail=f"文档不存在: doc_id={doc_id}"
+            )
+
+        # 获取第一个匹配的对象
+        object_name = objects[0]
+        logger.info(f"找到MinIO对象: {object_name}")
+
+        # 获取文件数据
+        file_data = minio.get_object_data(object_name)
+        if file_data is None:
+            raise HTTPException(
+                status_code=500,
+                detail=f"获取文档数据失败"
+            )
+
+        # 解析原始文件名
+        original_name = object_name.split("/", 1)[1] if "/" in object_name else object_name
+
+        # 获取Content-Type
+        content_type = minio._get_content_type(original_name)
+
+        logger.success(f"文档下载成功: {original_name}, 大小={len(file_data)}")
+
+        # 返回文件流（URL编码文件名以支持中文）
+        encoded_name = quote(original_name)
+        return StreamingResponse(
+            BytesIO(file_data),
+            media_type=content_type,
+            headers={
+                "Content-Disposition": f"attachment; filename*=UTF-8''{encoded_name}"
+            }
+        )
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"文档下载失败: {e}")
+        raise HTTPException(
+            status_code=500,
+            detail=f"文档下载失败: {str(e)}"
+        )
+
+
+@router.get("/list")
+async def list_documents():
+    """
+    列出所有已上传的文档（从MinIO获取）
+    """
+    try:
+        documents = _build_document_records()
+        return {"documents": documents, "total": len(documents)}
+
+    except Exception as e:
+        logger.error(f"列出文档失败: {e}")
+        return {"documents": [], "total": 0, "error": str(e)}
+
+
+@router.get("/management-list")
+async def get_document_management_list():
+    """
+    文档管理清单接口：仅返回最近的10条文档。
+    """
+    try:
+        documents = _build_document_records(limit=10)
+        return {"documents": documents, "total": len(documents), "limit": 10}
+
+    except Exception as e:
+        logger.error(f"获取文档管理清单失败: {e}")
+        return {"documents": [], "total": 0, "limit": 10, "error": str(e)}

backend/app/api/routes/knowledge.py

@@ -0,0 +1,81 @@
+# src/api/routes/knowledge.py
+"""知识库检索接口"""
+
+from fastapi import APIRouter, HTTPException
+from loguru import logger
+
+from ..models import SearchRequest, SearchResponse, SearchResultItem, ErrorResponse
+from app.services.document_processor import DocumentProcessor
+
+router = APIRouter(prefix="/knowledge", tags=["knowledge"])
+
+
+@router.post("/search", response_model=SearchResponse)
+async def search_knowledge(request: SearchRequest):
+    """
+    检索法规知识库
+
+    使用混合检索：Dense向量 + Sparse向量 + RRF融合
+
+    Args:
+        request: 检索请求参数
+    """
+    if not request.query or len(request.query.strip()) == 0:
+        raise HTTPException(
+            status_code=400,
+            detail="查询文本不能为空"
+        )
+
+    logger.info(f"收到检索请求: {request.query}")
+
+    try:
+        # 执行检索
+        processor = DocumentProcessor()
+        results = processor.search(
+            query=request.query,
+            top_k=request.top_k,
+            filters=request.filters
+        )
+        processor.close()
+
+        # 转换结果格式
+        result_items = []
+        for r in results:
+            item = SearchResultItem(
+                id=r.get("id", 0),
+                content=r.get("content", ""),
+                score=r.get("score", 0.0),
+                metadata=r.get("metadata", {})
+            )
+            result_items.append(item)
+
+        return SearchResponse(
+            query=request.query,
+            total=len(result_items),
+            results=result_items
+        )
+
+    except Exception as e:
+        logger.error(f"检索失败: {e}")
+        raise HTTPException(
+            status_code=500,
+            detail=f"检索失败: {str(e)}"
+        )
+
+
+@router.post("/retrieval", response_model=SearchResponse)
+async def knowledge_retrieval(request: SearchRequest):
+    """
+    知识检索接口（与架构文档对齐）
+
+    该接口实现完整的检索流程：
+    1. 意图识别
+    2. BM25关键词检索 + 向量语义检索（双路召回）
+    3. Cross-Encoder精排
+    4. 返回结果
+
+    Args:
+        request: 检索请求
+    """
+    # 当前版本使用混合检索，后续可添加精排步骤
+    return await search_knowledge(request)

backend/app/api/routes/rag.py

@@ -0,0 +1,74 @@
+from fastapi import APIRouter
+from sse_starlette.sse import EventSourceResponse
+from app.schemas.rag import RagChatRequest, QuickQuestionsResponse, QuickQuestion
+from app.services.mock_data import (
+    get_mock_quick_questions,
+    get_mock_retrieval,
+    get_mock_rag_answer,
+)
+import json
+import asyncio
+
+router = APIRouter(prefix="/rag", tags=["RAG问答"])
+
+
+@router.post("/chat")
+async def rag_chat(request: RagChatRequest):
+    """SSE流式问答"""
+
+    async def generate():
+        # 发送检索开始事件
+        yield {"event": "message", "data": json.dumps({"type": "retrieving"})}
+
+        # 模拟检索延迟
+        await asyncio.sleep(0.3)
+
+        # 执行检索
+        docs = get_mock_retrieval(request.query, top_k=request.top_k)
+
+        retrieved_data = [
+            {
+                "id": d["id"],
+                "score": d["score"],
+                "preview": d["preview"],
+                "doc_name": d.get("doc_name", ""),
+                "clause": d.get("clause", ""),
+            }
+            for d in docs
+        ]
+        yield {"event": "message", "data": json.dumps({"type": "retrieved", "docs": retrieved_data})}
+
+        # 发送生成开始事件
+        yield {"event": "message", "data": json.dumps({"type": "generating", "text": "正在生成答案..."})}
+
+        # 模拟生成延迟
+        await asyncio.sleep(0.2)
+
+        # 获取预设答案
+        answer = get_mock_rag_answer(request.query)
+
+        # 流式输出答案（按句子分割）
+        sentences = answer.split("\n\n")
+        for sentence in sentences:
+            if sentence.strip():
+                # 进一步分割长句子
+                chunks = sentence.split("\n")
+                for chunk in chunks:
+                    if chunk.strip():
+                        await asyncio.sleep(0.05)  # 模拟生成延迟
+                        yield {"event": "message", "data": json.dumps({"type": "chunk", "text": chunk + "\n"})}
+
+        # 发送完成事件
+        yield {"event": "message", "data": json.dumps({"type": "done"})}
+
+    return EventSourceResponse(generate())
+
+
+@router.get("/quick-questions", response_model=QuickQuestionsResponse)
+async def get_quick_questions():
+    """获取预设快捷问题"""
+    questions = [
+        QuickQuestion(id=q["id"], question=q["question"], category=q["category"])
+        for q in get_mock_quick_questions()
+    ]
+    return QuickQuestionsResponse(questions=questions)

backend/app/api/routes/status.py

@@ -0,0 +1,28 @@
+from fastapi import APIRouter
+from app.core.config import settings
+from app.services.mock_data import MOCK_SYSTEM_STATS, MOCK_SYSTEM_CONFIG
+
+router = APIRouter(prefix="/status", tags=["系统状态"])
+
+
+@router.get("/stats")
+async def get_stats():
+    """获取系统统计"""
+    # 返回预设统计数据
+    return MOCK_SYSTEM_STATS
+
+
+@router.get("/config")
+async def get_config():
+    """获取当前配置"""
+    return MOCK_SYSTEM_CONFIG
+
+
+@router.get("/milvus/health")
+async def milvus_health():
+    """Milvus健康检查"""
+    # 模拟连接状态（假数据模式下始终返回连接成功）
+    return {
+        "connected": True,
+        "collections": ["vehicle_regulations"],
+    }

backend/app/config/__init__.py

@@ -0,0 +1,6 @@
+# src/config/__init__.py
+"""配置模块"""
+
+from .settings import Settings, get_settings, settings
+
+__all__ = ["Settings", "get_settings", "settings"]

backend/app/config/logging.py

@@ -0,0 +1,32 @@
+# src/config/logging.py
+"""日志配置"""
+
+from loguru import logger
+import sys
+
+
+def setup_logging(level: str = "INFO"):
+    """设置日志配置"""
+
+    # 移除默认handler
+    logger.remove()
+
+    # 添加控制台输出
+    logger.add(
+        sys.stdout,
+        level=level,
+        format="<green>{time:YYYY-MM-DD HH:mm:ss}</green> | <level>{level: <8}</level> | <cyan>{name}</cyan>:<cyan>{function}</cyan>:<cyan>{line}</cyan> - <level>{message}</level>",
+        colorize=True
+    )
+
+    # 添加文件输出
+    logger.add(
+        "logs/app_{time:YYYY-MM-DD}.log",
+        level=level,
+        format="{time:YYYY-MM-DD HH:mm:ss} | {level: <8} | {name}:{function}:{line} - {message}",
+        rotation="00:00",
+        retention="7 days",
+        compression="zip"
+    )
+
+    return logger

backend/app/config/settings.py

@@ -0,0 +1,95 @@
+# src/config/settings.py
+"""配置管理 - 环境变量和默认配置"""
+
+from pydantic_settings import BaseSettings
+from pydantic import Field
+from typing import Optional
+from functools import lru_cache
+
+
+class Settings(BaseSettings):
+    """应用配置"""
+
+    # 应用基础配置
+    app_name: str = Field(default="AI Regulations Demo", description="Application name")
+    app_version: str = Field(default="0.1.0", description="应用版本")
+    debug: bool = Field(default=False, description="调试模式")
+
+    # Milvus向量数据库配置
+    milvus_host: str = Field(default="localhost", description="Milvus服务地址")
+    milvus_port: int = Field(default=19530, description="Milvus服务端口")
+    milvus_collection: str = Field(default="regulations", description="法规向量集合名称")
+    milvus_db_name: str = Field(default="default", description="Milvus数据库名称")
+
+    # 嵌入模型配置
+    embedding_model: str = Field(default="BAAI/bge-m3", description="嵌入模型名称")
+    embedding_dim: int = Field(default=1024, description="嵌入向量维度")
+    embedding_max_length: int = Field(default=8192, description="最大嵌入长度")
+    embedding_batch_size: int = Field(default=12, description="嵌入批处理大小")
+    embedding_use_fp16: bool = Field(default=True, description="使用FP16加速")
+
+    # MinIO对象存储配置
+    minio_endpoint: str = Field(default="localhost:9000", description="MinIO服务地址")
+    minio_access_key: str = Field(default="minioadmin", description="MinIO访问密钥")
+    minio_secret_key: str = Field(default="minioadmin123", description="MinIO秘密密钥")
+    minio_bucket: str = Field(default="upload-files", description="文档存储桶名称")
+    minio_secure: bool = Field(default=False, description="是否使用HTTPS")
+
+    # Redis配置
+    redis_host: str = Field(default="localhost", description="Redis服务地址")
+    redis_port: int = Field(default=6379, description="Redis服务端口")
+    redis_password: str = Field(default="", description="Redis密码")
+    redis_db: int = Field(default=0, description="Redis数据库编号")
+
+    # PostgreSQL配置
+    postgres_host: str = Field(default="localhost", description="PostgreSQL服务地址")
+    postgres_port: int = Field(default=5432, description="PostgreSQL服务端口")
+    postgres_user: str = Field(default="compliance", description="PostgreSQL用户名")
+    postgres_password: str = Field(default="compliance123", description="PostgreSQL密码")
+    postgres_db: str = Field(default="compliance_db", description="PostgreSQL数据库名称")
+
+    # 文档处理配置
+    chunk_size: int = Field(default=512, description="分块大小（字符数）")
+    chunk_overlap: int = Field(default=50, description="分块重叠大小")
+    max_file_size_mb: int = Field(default=100, description="最大文件大小(MB)")
+
+    # API配置
+    api_host: str = Field(default="0.0.0.0", description="API服务地址")
+    api_port: int = Field(default=8000, description="API服务端口")
+
+    # LLM配置
+    llm_provider: str = Field(default="deepseek", description="LLM提供商 (deepseek/qwen/qwen_vl)")
+    llm_model: str = Field(default="deepseek-v4-flash", description="LLM模型名称")
+    llm_max_tokens: int = Field(default=4096, description="LLM最大输出token数")
+    llm_temperature: float = Field(default=0.7, description="LLM温度参数")
+
+    # DeepSeek配置
+    deepseek_api_key: str = Field(default="", description="DeepSeek API密钥")
+    deepseek_base_url: str = Field(default="http://6.86.80.4:30080/v1", description="DeepSeek API地址")
+    deepseek_model: str = Field(default="deepseek-v4-flash", description="DeepSeek模型")
+
+    # Qwen配置（通过统一代理API）
+    qwen_api_key: str = Field(default="", description="Qwen API密钥")
+    qwen_base_url: str = Field(default="http://6.86.80.4:30080/v1", description="Qwen API地址")
+    qwen_model: str = Field(default="qwen3.5-flash", description="Qwen文本模型")
+    qwen_vl_model: str = Field(default="qwen3-vl-plus", description="Qwen视觉模型")
+
+    # RAG配置
+    rag_top_k: int = Field(default=5, description="检索召回数量")
+    rag_max_context_tokens: int = Field(default=2000, description="RAG最大上下文token数")
+    rag_summary_max_tokens: int = Field(default=10240, description="文档摘要最大token数")
+
+    class Config:
+        env_file = ".env"
+        env_file_encoding = "utf-8"
+        extra = "ignore"
+
+
+@lru_cache
+def get_settings() -> Settings:
+    """获取配置实例（缓存）"""
+    return Settings()
+
+
+# 导出默认配置实例
+settings = get_settings()

backend/app/core/__init__.py

@@ -0,0 +1,3 @@
+from .config import settings, Settings
+
+__all__ = ["settings", "Settings"]

backend/app/core/config.py

@@ -0,0 +1,41 @@
+from pydantic_settings import BaseSettings
+from typing import Optional
+
+
+class Settings(BaseSettings):
+    # DashScope API
+    dashscope_api_key: str = ""
+
+    # Milvus
+    milvus_host: str = "localhost"
+    milvus_port: int = 19530
+
+    # LLM配置
+    llm_model: str = "qwen-max"
+    embedding_model: str = "text-embedding-v3"
+    embedding_dim: int = 1536
+
+    # 检索配置
+    vector_top_k: int = 10
+    bm25_top_k: int = 10
+    final_top_k: int = 5
+
+    # 分块配置
+    chunk_size: int = 800
+    chunk_overlap: int = 50
+
+    # 服务配置
+    api_host: str = "0.0.0.0"
+    api_port: int = 8000
+
+    # Collection名称
+    regulations_collection: str = "vehicle_regulations"
+    compliance_collection: str = "compliance_cache"
+
+    class Config:
+        env_file = ".env"
+        env_file_encoding = "utf-8"
+        case_sensitive = False
+
+
+settings = Settings()

backend/app/main.py

@@ -0,0 +1,5 @@
+"""Backend application entrypoint."""
+
+from app.api.main import app
+
+__all__ = ["app"]

backend/app/schemas/__init__.py

@@ -0,0 +1,49 @@
+from .doc import (
+    DocumentUploadResponse,
+    DocumentInfo,
+    DocumentListResponse,
+    ChunkInfo,
+    ParseResponse,
+    EmbedResponse,
+)
+from .rag import (
+    RagChatRequest,
+    RetrievedDoc,
+    SourceInfo,
+    QuickQuestion,
+    QuickQuestionsResponse,
+)
+from .compliance import (
+    RiskLevel,
+    ComplianceStatus,
+    Regulation,
+    ComplianceSegment,
+    RiskDashboard,
+    PriorityAction,
+    ComplianceResult,
+    ComplianceChatRequest,
+    AnalyzeResponse,
+)
+
+__all__ = [
+    "DocumentUploadResponse",
+    "DocumentInfo",
+    "DocumentListResponse",
+    "ChunkInfo",
+    "ParseResponse",
+    "EmbedResponse",
+    "RagChatRequest",
+    "RetrievedDoc",
+    "SourceInfo",
+    "QuickQuestion",
+    "QuickQuestionsResponse",
+    "RiskLevel",
+    "ComplianceStatus",
+    "Regulation",
+    "ComplianceSegment",
+    "RiskDashboard",
+    "PriorityAction",
+    "ComplianceResult",
+    "ComplianceChatRequest",
+    "AnalyzeResponse",
+]

backend/app/schemas/compliance.py

@@ -0,0 +1,69 @@
+from pydantic import BaseModel
+from typing import Optional
+from enum import Enum
+
+
+class RiskLevel(str, Enum):
+    high = "high"
+    medium = "medium"
+    low = "low"
+
+
+class ComplianceStatus(str, Enum):
+    pass_status = "pass"
+    warning = "warning"
+    fail = "fail"
+
+
+class Regulation(BaseModel):
+    id: int
+    name: str
+    clause: Optional[str] = None
+    score: float
+    match_keyword: str
+    category: RiskLevel
+    full_content: str
+
+
+class ComplianceSegment(BaseModel):
+    id: int
+    index: int
+    intent: str
+    start_pos: int
+    end_pos: int
+    content: str
+    risk_level: RiskLevel
+    regulations: list[Regulation]
+
+
+class RiskDashboard(BaseModel):
+    score: float
+    high_risk_count: int
+    medium_risk_count: int
+    low_risk_count: int
+    need_fix_segments: int
+    status: ComplianceStatus
+    status_label: str
+
+
+class PriorityAction(BaseModel):
+    regulation: str
+    issue: str
+    suggestion: str
+    severity: RiskLevel
+
+
+class ComplianceResult(BaseModel):
+    task_id: str
+    dashboard: RiskDashboard
+    segments: list[ComplianceSegment]
+    priority_actions: list[PriorityAction]
+
+
+class ComplianceChatRequest(BaseModel):
+    query: str
+
+
+class AnalyzeResponse(BaseModel):
+    task_id: str
+    status: str = "processing"

backend/app/schemas/doc.py

@@ -0,0 +1,44 @@
+from pydantic import BaseModel
+from typing import Optional
+from datetime import datetime
+
+
+class DocumentUploadResponse(BaseModel):
+    doc_id: str
+    filename: str
+    size: int
+    status: str = "uploaded"
+
+
+class DocumentInfo(BaseModel):
+    id: str
+    name: str
+    chunks: int
+    status: str
+    created_at: Optional[datetime] = None
+
+
+class DocumentListResponse(BaseModel):
+    docs: list[DocumentInfo]
+
+
+class ChunkInfo(BaseModel):
+    chunk_id: str
+    doc_name: str
+    clause_id: Optional[str] = None
+    chapter: Optional[str] = None
+    content: str
+    token_count: int
+    chunk_index: int
+
+
+class ParseResponse(BaseModel):
+    doc_id: str
+    chunks: int
+    status: str = "parsed"
+
+
+class EmbedResponse(BaseModel):
+    doc_id: str
+    vectors: int
+    status: str = "embedded"

backend/app/schemas/rag.py

@@ -0,0 +1,31 @@
+from pydantic import BaseModel
+from typing import Optional
+
+
+class RagChatRequest(BaseModel):
+    query: str
+    top_k: int = 5
+
+
+class RetrievedDoc(BaseModel):
+    id: str
+    doc_name: str
+    clause_id: Optional[str] = None
+    score: float
+    content: str
+    preview: str
+
+
+class SourceInfo(BaseModel):
+    name: str
+    clause: Optional[str] = None
+
+
+class QuickQuestion(BaseModel):
+    id: str
+    question: str
+    category: str
+
+
+class QuickQuestionsResponse(BaseModel):
+    questions: list[QuickQuestion]

backend/app/services/__init__.py

@@ -0,0 +1,3 @@
+"""Backend service package."""
+
+__all__: list[str] = []

backend/app/services/agent/__init__.py

@@ -0,0 +1,7 @@
+# src/services/agent/__init__.py
+"""Agent服务模块"""
+
+from .qa_agent import QAAgent, ask_compliance_question
+from .session_manager import SessionManager, ChatSession
+
+__all__ = ["QAAgent", "ask_compliance_question", "SessionManager", "ChatSession"]

backend/app/services/agent/qa_agent.py

@@ -0,0 +1,412 @@
+# src/services/agent/qa_agent.py
+"""RAG问答Agent - 合规智能问答核心实现"""
+
+import time
+from typing import List, Dict, Optional, Any, Generator
+from dataclasses import dataclass, field
+from loguru import logger
+
+from app.services.llm import get_llm_client, BaseLLMClient, LLMResponse
+from app.services.llm.llm_factory import LLMFactory
+from app.services.rag.retriever import Retriever, RetrievedDocument
+from app.services.rag.context_builder import ContextBuilder, RAGContext
+from app.services.rag.prompt_templates import get_prompt_template, PromptTemplate
+from app.config.settings import settings
+
+
+@dataclass
+class AgentResponse:
+    """Agent响应结果"""
+    answer: str
+    sources: List[Dict] = field(default_factory=list)
+    model: str = ""
+    latency_ms: int = 0
+    retrieved_count: int = 0
+    context_tokens: int = 0
+    truncated: bool = False
+    error: Optional[str] = None
+
+    @property
+    def is_success(self) -> bool:
+        return self.error is None
+
+
+@dataclass
+class AgentConfig:
+    """Agent配置"""
+    llm_provider: str = "deepseek"
+    llm_model: str = "deepseek-v4-flash"
+    top_k: int = 5
+    min_score: float = 0.3
+    max_context_tokens: int = 2000
+    temperature: float = 0.7
+    prompt_template: str = "compliance_qa"
+    include_metadata: bool = True
+
+
+class QAAgent:
+    """
+    合规问答Agent
+
+    核心流程：
+    1. 接收用户问题
+    2. Milvus混合检索相关法规条款
+    3. 构建RAG上下文
+    4. 调用LLM生成回答
+    5. 返回答案和引用来源
+
+    使用示例：
+        agent = QAAgent()
+        response = agent.ask("机动车安全技术检验有哪些要求？")
+        print(response.answer)
+        for source in response.sources:
+            print(f"引用: {source['doc_name']} - {source['clause_number']}")
+    """
+
+    def __init__(self, config: Optional[AgentConfig] = None):
+        """
+        初始化问答Agent
+
+        Args:
+            config: Agent配置（可选，使用默认配置）
+        """
+        self.config = config or AgentConfig(
+            llm_provider=settings.llm_provider,
+            llm_model=settings.llm_model,
+            top_k=settings.rag_top_k,
+            max_context_tokens=settings.rag_max_context_tokens
+        )
+
+        # 初始化组件（延迟加载）
+        self.llm: Optional[BaseLLMClient] = None
+        self.retriever: Optional[Retriever] = None
+        self.context_builder: Optional[ContextBuilder] = None
+
+        logger.info(f"问答Agent初始化: provider={self.config.llm_provider}, model={self.config.llm_model}")
+
+    def _init_llm(self):
+        """延迟初始化LLM客户端（优先使用全局缓存）"""
+        if self.llm is None:
+            # 尝试先获取全局缓存的客户端
+            cached = LLMFactory.get_global_client(self.config.llm_provider, self.config.llm_model)
+            if cached:
+                self.llm = cached
+                logger.debug(f"使用全局缓存的LLM客户端: {self.config.llm_provider} - {self.config.llm_model}")
+            else:
+                logger.info("创建新的LLM客户端...")
+                self.llm = get_llm_client(
+                    provider=self.config.llm_provider,
+                    model=self.config.llm_model,
+                    temperature=self.config.temperature
+                )
+
+    def _init_retriever(self):
+        """延迟初始化检索器"""
+        if self.retriever is None:
+            logger.info("初始化检索器...")
+            self.retriever = Retriever(
+                top_k=self.config.top_k,
+                min_score=self.config.min_score
+            )
+
+    def _init_context_builder(self):
+        """延迟初始化上下文构建器"""
+        if self.context_builder is None:
+            logger.info("初始化上下文构建器...")
+            self.context_builder = ContextBuilder(
+                max_context_tokens=self.config.max_context_tokens,
+                include_metadata=self.config.include_metadata
+            )
+
+    def ask(
+        self,
+        query: str,
+        filters: Optional[str] = None,
+        prompt_template: Optional[str] = None
+    ) -> AgentResponse:
+        """
+        回答用户问题
+
+        Args:
+            query: 用户问题
+            filters: 检索过滤条件（如 "regulation_type=='车辆安全'"）
+            prompt_template: Prompt模板名称（可选，覆盖默认配置）
+
+        Returns:
+            AgentResponse: 包含答案和引用来源的响应对象
+        """
+        start_time = time.time()
+        logger.info(f"收到问题: {query}")
+
+        try:
+            # Step 1: 检索相关法规
+            self._init_retriever()
+            documents = self.retriever.retrieve(query, filters)
+            retrieved_count = len(documents)
+
+            if retrieved_count == 0:
+                return AgentResponse(
+                    answer="抱歉，未找到与您的问题相关的法规条款。请尝试用不同的关键词重新提问，或提供更具体的法规名称。",
+                    retrieved_count=0,
+                    error="no_retrieved_documents"
+                )
+
+            # Step 2: 构建RAG上下文
+            self._init_context_builder()
+            template_name = prompt_template or self.config.prompt_template
+            template = get_prompt_template(template_name)
+
+            context = self.context_builder.build(
+                query=query,
+                documents=documents,
+                system_prompt=template.system_prompt
+            )
+
+            # Step 3: 构建LLM输入消息
+            messages = self._build_messages(template, context)
+
+            # Step 4: 调用LLM生成回答
+            self._init_llm()
+            llm_response = self.llm.chat(
+                messages=messages,
+                temperature=self.config.temperature
+            )
+
+            if not llm_response.is_success:
+                return AgentResponse(
+                    answer="",
+                    retrieved_count=retrieved_count,
+                    error=llm_response.error
+                )
+
+            latency_ms = int((time.time() - start_time) * 1000)
+
+            # Step 5: 返回结果
+            logger.success(f"问答完成: {latency_ms}ms, {retrieved_count}条引用")
+
+            return AgentResponse(
+                answer=llm_response.content,
+                sources=context.sources,
+                model=llm_response.model,
+                latency_ms=latency_ms,
+                retrieved_count=retrieved_count,
+                context_tokens=context.total_tokens,
+                truncated=context.truncated
+            )
+
+        except Exception as e:
+            logger.error(f"问答失败: {e}")
+            return AgentResponse(
+                answer="",
+                error=str(e)
+            )
+
+    def ask_with_context(
+        self,
+        query: str,
+        documents: List[RetrievedDocument],
+        prompt_template: Optional[str] = None
+    ) -> AgentResponse:
+        """
+        使用提供的文档回答问题（不执行检索）
+
+        Args:
+            query: 用户问题
+            documents: 已检索的文档列表
+            prompt_template: Prompt模板名称
+
+        Returns:
+            AgentResponse: 响应结果
+        """
+        start_time = time.time()
+
+        try:
+            self._init_context_builder()
+            self._init_llm()
+
+            template_name = prompt_template or self.config.prompt_template
+            template = get_prompt_template(template_name)
+
+            context = self.context_builder.build(
+                query=query,
+                documents=documents,
+                system_prompt=template.system_prompt
+            )
+
+            messages = self._build_messages(template, context)
+
+            llm_response = self.llm.chat(messages)
+
+            latency_ms = int((time.time() - start_time) * 1000)
+
+            return AgentResponse(
+                answer=llm_response.content,
+                sources=context.sources,
+                model=llm_response.model,
+                latency_ms=latency_ms,
+                retrieved_count=len(documents),
+                context_tokens=context.total_tokens,
+                truncated=context.truncated
+            )
+
+        except Exception as e:
+            logger.error(f"问答失败: {e}")
+            return AgentResponse(answer="", error=str(e))
+
+    def _build_messages(
+        self,
+        template: PromptTemplate,
+        context: RAGContext
+    ) -> List[Dict[str, str]]:
+        """构建LLM输入消息"""
+        user_content = template.user_template.format(
+            context=context.context_text,
+            query=context.user_query
+        )
+
+        return [
+            {"role": "system", "content": template.system_prompt},
+            {"role": "user", "content": user_content}
+        ]
+
+    def ask_stream(
+        self,
+        query: str,
+        filters: Optional[str] = None,
+        prompt_template: Optional[str] = None
+    ) -> Generator[Dict[str, Any], None, None]:
+        """
+        流式回答用户问题（SSE模式）
+
+        返回事件类型:
+        - {"event": "status", "data": "正在检索..."}  - 状态更新
+        - {"event": "sources", "data": [...]}  - 引用来源
+        - {"event": "content", "data": "文本片段"}  - 回答内容
+        - {"event": "done", "data": {"latency_ms": ..., "model": ...}}  - 完成
+
+        Args:
+            query: 用户问题
+            filters: 检索过滤条件
+            prompt_template: Prompt模板名称
+
+        Yields:
+            Dict: SSE事件数据
+        """
+        start_time = time.time()
+        logger.info(f"收到流式问题: {query}")
+
+        try:
+            # Step 1: 检索相关法规
+            yield {"event": "status", "data": "正在检索相关法规..."}
+            self._init_retriever()
+            documents = self.retriever.retrieve(query, filters)
+            retrieved_count = len(documents)
+
+            if retrieved_count == 0:
+                yield {"event": "status", "data": "未找到相关法规"}
+                yield {"event": "content", "data": "抱歉，未找到与您的问题相关的法规条款。请尝试用不同的关键词重新提问。"}
+                yield {"event": "done", "data": {"latency_ms": 0, "retrieved_count": 0}}
+                return
+
+            # Step 2: 发送检索结果
+            yield {"event": "status", "data": f"找到{retrieved_count}条相关法规，正在生成回答..."}
+            sources = [
+                {
+                    "doc_name": doc.doc_name,
+                    "doc_id": doc.doc_id,
+                    "clause_number": doc.clause_number,
+                    "score": doc.score
+                }
+                for doc in documents[:5]  # 只返回前5条引用
+            ]
+            yield {"event": "sources", "data": sources}
+
+            # Step 3: 构建RAG上下文
+            self._init_context_builder()
+            template_name = prompt_template or self.config.prompt_template
+            template = get_prompt_template(template_name)
+            context = self.context_builder.build(
+                query=query,
+                documents=documents,
+                system_prompt=template.system_prompt
+            )
+
+            # Step 4: 构建LLM输入消息
+            messages = self._build_messages(template, context)
+
+            # Step 5: 流式调用LLM生成回答
+            self._init_llm()
+            full_answer = ""
+
+            # 检查LLM是否支持流式输出
+            if hasattr(self.llm, 'stream_chat'):
+                yield {"event": "status", "data": "思考中..."}
+                for chunk in self.llm.stream_chat(
+                    messages=messages,
+                    temperature=self.config.temperature
+                ):
+                    full_answer += chunk
+                    yield {"event": "content", "data": chunk}
+            else:
+                # 如果不支持流式，回退到普通调用
+                yield {"event": "status", "data": "生成回答中..."}
+                llm_response = self.llm.chat(
+                    messages=messages,
+                    temperature=self.config.temperature
+                )
+                if llm_response.is_success:
+                    full_answer = llm_response.content
+                    yield {"event": "content", "data": full_answer}
+
+            # Step 6: 发送完成事件
+            latency_ms = int((time.time() - start_time) * 1000)
+            logger.success(f"流式问答完成: {latency_ms}ms, {retrieved_count}条引用")
+
+            yield {
+                "event": "done",
+                "data": {
+                    "latency_ms": latency_ms,
+                    "model": self.config.llm_model,
+                    "retrieved_count": retrieved_count,
+                    "context_tokens": context.total_tokens
+                }
+            }
+
+        except Exception as e:
+            logger.error(f"流式问答失败: {e}")
+            yield {"event": "error", "data": str(e)}
+
+    def close(self):
+        """关闭Agent资源（不关闭LLM客户端，因为它全局缓存）"""
+        if self.retriever:
+            self.retriever.close()
+        logger.info("问答Agent已关闭")
+
+
+def ask_compliance_question(
+    query: str,
+    provider: str = "deepseek",
+    model: str = "deepseek-v4-flash",
+    top_k: int = 10
+) -> AgentResponse:
+    """
+    便捷函数：问答合规问题
+
+    Args:
+        query: 用户问题
+        provider: LLM提供商
+        model: LLM模型
+        top_k: 检索数量
+
+    Returns:
+        AgentResponse: 响应结果
+    """
+    config = AgentConfig(
+        llm_provider=provider,
+        llm_model=model,
+        top_k=top_k
+    )
+    agent = QAAgent(config)
+    response = agent.ask(query)
+    agent.close()
+    return response

backend/app/services/agent/session_manager.py

@@ -0,0 +1,247 @@
+# src/services/agent/session_manager.py
+"""多轮对话会话管理"""
+
+import time
+import uuid
+from typing import Dict, List, Optional
+from dataclasses import dataclass, field
+from loguru import logger
+
+
+@dataclass
+class ChatMessage:
+    """对话消息"""
+    role: str  # "user" / "assistant" / "system"
+    content: str
+    timestamp: int
+    sources: List[Dict] = field(default_factory=list)
+    metadata: Dict = field(default_factory=dict)
+
+
+@dataclass
+class ChatSession:
+    """对话会话"""
+    session_id: str
+    messages: List[ChatMessage] = field(default_factory=list)
+    created_at: int = field(default_factory=lambda: int(time.time()))
+    updated_at: int = field(default_factory=lambda: int(time.time()))
+    metadata: Dict = field(default_factory=dict)
+
+    def add_user_message(self, content: str) -> ChatMessage:
+        """添加用户消息"""
+        message = ChatMessage(
+            role="user",
+            content=content,
+            timestamp=int(time.time())
+        )
+        self.messages.append(message)
+        self.updated_at = int(time.time())
+        return message
+
+    def add_assistant_message(
+        self,
+        content: str,
+        sources: List[Dict] = None
+    ) -> ChatMessage:
+        """添加助手消息"""
+        message = ChatMessage(
+            role="assistant",
+            content=content,
+            timestamp=int(time.time()),
+            sources=sources or []
+        )
+        self.messages.append(message)
+        self.updated_at = int(time.time())
+        return message
+
+    def get_history(self, max_turns: int = 5) -> List[Dict[str, str]]:
+        """获取历史对话（用于LLM上下文）"""
+        history = []
+        # 获取最近N轮对话（每轮包含user + assistant）
+        recent_messages = self.messages[-(max_turns * 2):]
+
+        for msg in recent_messages:
+            history.append({
+                "role": msg.role,
+                "content": msg.content
+            })
+
+        return history
+
+    def clear_history(self):
+        """清空对话历史"""
+        self.messages = []
+        self.updated_at = int(time.time())
+        logger.info(f"会话历史已清空: {self.session_id}")
+
+    @property
+    def message_count(self) -> int:
+        """消息数量"""
+        return len(self.messages)
+
+    @property
+    def is_empty(self) -> bool:
+        """是否为空会话"""
+        return len(self.messages) == 0
+
+
+class SessionManager:
+    """
+    会话管理器
+
+    功能：
+    - 创建/获取/删除会话
+    - 会话超时清理
+    - 会话历史记录管理
+
+    使用示例：
+        manager = SessionManager()
+
+        # 创建会话
+        session = manager.create_session()
+
+        # 添加消息
+        session.add_user_message("什么是机动车安全技术检验？")
+        session.add_assistant_message("根据GB 7258...", sources=[...])
+
+        # 获取历史（用于LLM多轮对话）
+        history = session.get_history(max_turns=3)
+    """
+
+    def __init__(
+        self,
+        max_sessions: int = 100,
+        session_timeout_minutes: int = 30
+    ):
+        """
+        初始化会话管理器
+
+        Args:
+            max_sessions: 最大会话数量
+            session_timeout_minutes: 会话超时时间（分钟）
+        """
+        self.max_sessions = max_sessions
+        self.session_timeout = session_timeout_minutes * 60
+
+        # 会话存储（内存）
+        self._sessions: Dict[str, ChatSession] = {}
+
+        logger.info(f"会话管理器初始化: max_sessions={max_sessions}, timeout={session_timeout_minutes}min")
+
+    def create_session(self, metadata: Dict = None) -> ChatSession:
+        """
+        创建新会话
+
+        Args:
+            metadata: 会话元数据（可选）
+
+        Returns:
+            ChatSession: 新创建的会话
+        """
+        # 检查会话数量限制
+        if len(self._sessions) >= self.max_sessions:
+            # 清理过期会话
+            self._cleanup_expired_sessions()
+
+            # 如果仍然超出限制，删除最老的会话
+            if len(self._sessions) >= self.max_sessions:
+                oldest_id = min(
+                    self._sessions.keys(),
+                    key=lambda x: self._sessions[x].created_at
+                )
+                self.delete_session(oldest_id)
+                logger.warning(f"删除最老会话以腾出空间: {oldest_id}")
+
+        session_id = str(uuid.uuid4())[:8]
+        session = ChatSession(
+            session_id=session_id,
+            metadata=metadata or {}
+        )
+
+        self._sessions[session_id] = session
+        logger.info(f"创建新会话: {session_id}")
+
+        return session
+
+    def get_session(self, session_id: str) -> Optional[ChatSession]:
+        """
+        获取会话
+
+        Args:
+            session_id: 会话ID
+
+        Returns:
+            ChatSession: 会话对象（如不存在返回None）
+        """
+        session = self._sessions.get(session_id)
+
+        if session:
+            # 检查是否过期
+            if self._is_session_expired(session):
+                self.delete_session(session_id)
+                logger.info(f"会话已过期，已删除: {session_id}")
+                return None
+
+        return session
+
+    def delete_session(self, session_id: str) -> bool:
+        """
+        删除会话
+
+        Args:
+            session_id: 会话ID
+
+        Returns:
+            bool: 是否成功删除
+        """
+        if session_id in self._sessions:
+            del self._sessions[session_id]
+            logger.info(f"删除会话: {session_id}")
+            return True
+        return False
+
+    def list_sessions(self) -> List[Dict]:
+        """
+        列出所有会话
+
+        Returns:
+            List[Dict]: 会话列表摘要
+        """
+        return [
+            {
+                "session_id": s.session_id,
+                "message_count": s.message_count,
+                "created_at": s.created_at,
+                "updated_at": s.updated_at
+            }
+            for s in self._sessions.values()
+        ]
+
+    def _is_session_expired(self, session: ChatSession) -> bool:
+        """检查会话是否过期"""
+        current_time = int(time.time())
+        return (current_time - session.updated_at) > self.session_timeout
+
+    def _cleanup_expired_sessions(self) -> int:
+        """清理过期会话"""
+        expired_ids = [
+            sid for sid, session in self._sessions.items()
+            if self._is_session_expired(session)
+        ]
+
+        for sid in expired_ids:
+            self.delete_session(sid)
+
+        if expired_ids:
+            logger.info(f"清理过期会话: {len(expired_ids)}个")
+
+        return len(expired_ids)
+
+    def get_session_count(self) -> int:
+        """获取当前会话数量"""
+        return len(self._sessions)
+
+    def clear_all_sessions(self):
+        """清空所有会话"""
+        self._sessions.clear()
+        logger.info("所有会话已清空")

backend/app/services/document_processor.py

@@ -0,0 +1,404 @@
+# src/services/document_processor.py
+"""文档处理主流程 - 解析→摘要→分块→嵌入→入库"""
+
+import os
+from typing import List, Dict, Optional
+from dataclasses import dataclass
+from loguru import logger
+import uuid
+
+from .parser.pdf_parser import PDFParser
+from .parser.docx_parser import DocxParser
+from .parser.mineru_parser import ParserOrchestrator
+from .embedding.text_chunker import RegulationChunker, TextChunk
+from .embedding.bge_m3_embedder import BGEM3Embedder, EmbeddingResult
+from .storage.milvus_client import MilvusClient
+from .llm.document_summarizer import DocumentSummarizer, DocumentSummary
+from app.config.settings import settings
+
+
+@dataclass
+class ProcessingResult:
+    """文档处理结果"""
+    doc_id: str
+    doc_name: str
+    success: bool
+    num_chunks: int = 0
+    message: str = ""
+    markdown_text: str = ""
+    summary: str = ""
+    summary_latency_ms: int = 0
+
+
+class DocumentProcessor:
+    """
+    文档处理服务 - 完整处理流程
+
+    流程：
+    1. 文档解析（PDF/DOCX → Markdown）
+    2. 智能分块（章节级+条款级）
+    3. LLM摘要生成（可选）
+    4. 向量嵌入（BGE-M3 Dense+Sparse）
+    5. 存储入库（Milvus向量数据库）
+    """
+
+    def __init__(
+        self,
+        chunk_size: int = None,
+        embedding_model: str = None,
+        use_mineru: bool = True,
+        generate_summary: bool = False,  # 默认不生成摘要，节省约60秒
+        llm_provider: str = None,
+        llm_model: str = None
+    ):
+        """
+        初始化文档处理器
+
+        Args:
+            chunk_size: 分块大小
+            embedding_model: 嵌入模型名称
+            use_mineru: 是否优先使用MinerU解析
+            generate_summary: 是否生成文档摘要（默认False，可节省约60秒处理时间）
+            llm_provider: LLM提供商
+            llm_model: LLM模型名称
+        """
+        self.chunk_size = chunk_size or settings.chunk_size
+        self.embedding_model = embedding_model or settings.embedding_model
+        self.use_mineru = use_mineru
+        self.generate_summary = generate_summary
+        self.llm_provider = llm_provider or settings.llm_provider
+        self.llm_model = llm_model or settings.llm_model
+
+        # 初始化各组件
+        logger.info("初始化文档处理组件...")
+
+        # 解析器
+        self.parser = ParserOrchestrator()
+
+        # 分块器
+        self.chunker = RegulationChunker(chunk_size=self.chunk_size)
+
+        # 嵌入模型（延迟加载）
+        self.embedder: Optional[BGEM3Embedder] = None
+
+        # Milvus客户端（延迟连接）
+        self.milvus: Optional[MilvusClient] = None
+
+        # 摘要生成器（延迟加载）
+        self.summarizer: Optional[DocumentSummarizer] = None
+
+        logger.success("文档处理器初始化完成")
+
+    def _init_embedder(self):
+        """延迟初始化嵌入模型"""
+        if self.embedder is None:
+            logger.info("加载嵌入模型...")
+            self.embedder = BGEM3Embedder(model_name=self.embedding_model)
+
+    def _init_milvus(self):
+        """延迟初始化Milvus连接"""
+        if self.milvus is None:
+            logger.info("连接Milvus...")
+            self.milvus = MilvusClient()
+            self.milvus.connect()
+            self.milvus.create_collection(recreate=False)
+            self.milvus.load_collection()
+
+    def _init_summarizer(self):
+        """延迟初始化摘要生成器"""
+        if self.summarizer is None:
+            logger.info("初始化摘要生成器...")
+            self.summarizer = DocumentSummarizer(
+                provider=self.llm_provider,
+                model=self.llm_model
+            )
+
+    def process(
+        self,
+        file_path: str,
+        doc_id: Optional[str] = None,
+        doc_name: Optional[str] = None,
+        regulation_type: str = "",
+        version: str = ""
+    ) -> ProcessingResult:
+        """
+        处理单个文档
+
+        Args:
+            file_path: 文档文件路径
+            doc_id: 文档ID（可选，默认自动生成）
+            doc_name: 文档名称（可选，默认从文件名获取）
+            regulation_type: 法规类型
+            version: 文档版本
+
+        Returns:
+            ProcessingResult: 处理结果
+        """
+        # 生成或使用传入的文档ID
+        if doc_id is None:
+            doc_id = str(uuid.uuid4())[:8]
+
+        # 获取文档名称
+        if doc_name is None:
+            doc_name = os.path.basename(file_path)
+
+        logger.info(f"开始处理文档: {doc_name} (ID: {doc_id})")
+
+        # 初始化结果变量
+        summary = ""
+        summary_latency_ms = 0
+
+        try:
+            # 1. 文档解析
+            logger.info("Step 1: 文档解析")
+            markdown_text = self._parse_document(file_path)
+
+            if not markdown_text:
+                return ProcessingResult(
+                    doc_id=doc_id,
+                    doc_name=doc_name,
+                    success=False,
+                    message="文档解析失败，内容为空"
+                )
+
+            # 2. LLM摘要生成（可选）
+            if self.generate_summary:
+                logger.info("Step 2: LLM摘要生成")
+                self._init_summarizer()
+                summary_result = self.summarizer.summarize(
+                    doc_name,
+                    markdown_text,
+                    regulation_type
+                )
+                if summary_result.is_success:
+                    summary = summary_result.summary
+                    summary_latency_ms = summary_result.latency_ms
+                    logger.success(f"摘要生成完成: {summary_latency_ms}ms")
+                else:
+                    logger.warning(f"摘要生成失败: {summary_result.error}")
+            else:
+                logger.info("Step 2: 跳过摘要生成（未勾选，节省约60秒）")
+
+            # 3. 智能分块
+            logger.info("Step 3: 智能分块")
+            chunks = self._chunk_document(
+                markdown_text,
+                doc_id,
+                doc_name,
+                regulation_type,
+                version
+            )
+
+            if not chunks:
+                return ProcessingResult(
+                    doc_id=doc_id,
+                    doc_name=doc_name,
+                    success=False,
+                    message="分块失败，无有效内容",
+                    markdown_text=markdown_text,
+                    summary=summary
+                )
+
+            # 4. 向量嵌入
+            logger.info("Step 4: 向量嵌入")
+            embeddings = self._embed_chunks(chunks)
+
+            if embeddings is None:
+                return ProcessingResult(
+                    doc_id=doc_id,
+                    doc_name=doc_name,
+                    success=False,
+                    message="向量嵌入失败",
+                    markdown_text=markdown_text,
+                    summary=summary
+                )
+
+            # 5. 存储入库
+            logger.info("Step 5: 存储入库")
+            inserted_ids = self._insert_to_milvus(chunks, embeddings)
+
+            logger.success(f"文档处理完成: {doc_name}, 共{len(inserted_ids)}条记录")
+
+            return ProcessingResult(
+                doc_id=doc_id,
+                doc_name=doc_name,
+                success=True,
+                num_chunks=len(inserted_ids),
+                message="处理成功",
+                markdown_text=markdown_text,
+                summary=summary,
+                summary_latency_ms=summary_latency_ms
+            )
+
+        except Exception as e:
+            logger.error(f"文档处理失败: {e}")
+            return ProcessingResult(
+                doc_id=doc_id,
+                doc_name=doc_name,
+                success=False,
+                message=f"处理失败: {str(e)}"
+            )
+
+    def _parse_document(self, file_path: str) -> str:
+        """解析文档"""
+        ext = os.path.splitext(file_path)[1].lower()
+
+        try:
+            if ext == ".pdf":
+                # PDF文档解析（优先MinerU，回退PyMuPDF）
+                markdown_text = self.parser.parse_pdf(file_path, prefer_mineru=self.use_mineru)
+            elif ext in [".docx", ".doc"]:
+                # Word文档解析
+                markdown_text = self.parser.parse_docx(file_path)
+            else:
+                logger.warning(f"不支持的文件类型: {ext}")
+                return ""
+
+            logger.success(f"文档解析完成，内容长度: {len(markdown_text)}字符")
+            return markdown_text
+
+        except Exception as e:
+            logger.error(f"文档解析失败: {e}")
+            return ""
+
+    def _chunk_document(
+        self,
+        markdown_text: str,
+        doc_id: str,
+        doc_name: str,
+        regulation_type: str,
+        version: str
+    ) -> List[TextChunk]:
+        """分块文档"""
+        try:
+            chunks = self.chunker.chunk_document(
+                markdown_text,
+                doc_id=doc_id,
+                doc_name=doc_name,
+                regulation_type=regulation_type,
+                version=version
+            )
+            logger.success(f"分块完成，共{len(chunks)}个chunk")
+            return chunks
+
+        except Exception as e:
+            logger.error(f"分块失败: {e}")
+            return []
+
+    def _embed_chunks(self, chunks: List[TextChunk]) -> Optional[EmbeddingResult]:
+        """嵌入分块"""
+        try:
+            # 延迟初始化嵌入模型
+            self._init_embedder()
+
+            # 提取文本内容
+            texts = [chunk.content for chunk in chunks]
+
+            # 执行嵌入
+            embeddings = self.embedder.embed(texts)
+
+            logger.success(f"嵌入完成，向量数: {len(embeddings.dense_embeddings)}")
+            return embeddings
+
+        except Exception as e:
+            logger.error(f"嵌入失败: {e}")
+            return None
+
+    def _insert_to_milvus(
+        self,
+        chunks: List[TextChunk],
+        embeddings: EmbeddingResult
+    ) -> List[int]:
+        """插入Milvus"""
+        try:
+            # 延迟初始化Milvus
+            self._init_milvus()
+
+            # 执行插入
+            inserted_ids = self.milvus.insert_chunks(chunks, embeddings)
+
+            logger.success(f"入库完成，共{len(inserted_ids)}条记录")
+            return inserted_ids
+
+        except Exception as e:
+            logger.error(f"入库失败: {e}")
+            return []
+
+    def search(
+        self,
+        query: str,
+        top_k: int = 10,
+        filters: Optional[str] = None
+    ) -> List[Dict]:
+        """
+        检索法规内容
+
+        Args:
+            query: 查询文本
+            top_k: 返回结果数量
+            filters: 过滤条件
+
+        Returns:
+            List[Dict]: 检索结果
+        """
+        logger.info(f"执行检索: {query}")
+
+        try:
+            # 延迟初始化
+            self._init_embedder()
+            self._init_milvus()
+
+            # 生成查询向量
+            query_embedding = self.embedder.embed_single(query)
+
+            # 执行混合检索
+            results = self.milvus.hybrid_search(
+                query_dense=query_embedding['dense'].tolist(),
+                query_sparse=query_embedding['sparse'],
+                top_k=top_k,
+                filters=filters
+            )
+
+            # 转换为字典格式
+            result_dicts = []
+            for r in results:
+                result_dicts.append({
+                    "id": r.id,
+                    "content": r.content,
+                    "score": r.score,
+                    "metadata": r.metadata
+                })
+
+            logger.success(f"检索完成，返回{len(result_dicts)}条结果")
+            return result_dicts
+
+        except Exception as e:
+            logger.error(f"检索失败: {e}")
+            return []
+
+    def close(self):
+        """关闭连接"""
+        if self.milvus:
+            self.milvus.disconnect()
+        logger.info("文档处理器已关闭")
+
+
+def process_document(
+    file_path: str,
+    doc_name: Optional[str] = None,
+    regulation_type: str = "",
+    version: str = ""
+) -> ProcessingResult:
+    """便捷函数：处理单个文档"""
+    processor = DocumentProcessor()
+    result = processor.process(file_path, doc_name, regulation_type, version)
+    processor.close()
+    return result
+
+
+def search_regulations(query: str, top_k: int = 10) -> List[Dict]:
+    """便捷函数：检索法规"""
+    processor = DocumentProcessor()
+    results = processor.search(query, top_k)
+    processor.close()
+    return results

backend/app/services/embedding/__init__.py

@@ -0,0 +1,7 @@
+# src/services/embedding/__init__.py
+"""嵌入和分块服务"""
+
+from .text_chunker import RegulationChunker
+from .bge_m3_embedder import BGEM3Embedder
+
+__all__ = ["RegulationChunker", "BGEM3Embedder"]

backend/app/services/embedding/bge_m3_embedder.py

@@ -0,0 +1,296 @@
+# src/services/embedding/bge_m3_embedder.py
+"""BGE-M3嵌入服务 - Dense+Sparse双路向量生成"""
+
+import numpy as np
+from typing import List, Dict, Optional, Union
+from dataclasses import dataclass, field
+from loguru import logger
+import torch
+import os
+
+# 设置HuggingFace镜像（国内网络）
+if 'HF_ENDPOINT' not in os.environ:
+    os.environ['HF_ENDPOINT'] = 'https://hf-mirror.com'
+
+# 本地模型路径（按优先级检查）
+LOCAL_MODEL_PATHS = [
+    os.path.expanduser("~/.cache/modelscope/Xorbits/bge-m3"),  # ModelScope下载路径
+    os.path.expanduser("~/.cache/huggingface/hub/models--BAAI--bge-m3/snapshots/main"),  # HuggingFace本地路径
+]
+
+
+@dataclass
+class EmbeddingResult:
+    """嵌入结果"""
+    dense_embeddings: np.ndarray  # Dense向量（语义检索）
+    sparse_embeddings: List[Dict[int, float]]  # Sparse向量（关键词匹配）
+    texts: List[str]
+    dim: int = 1024
+
+
+class BGEM3Embedder:
+    """
+    BGE-M3多语言嵌入模型服务
+
+    BGE-M3是BAAI发布的多语言嵌入模型，支持：
+    - Dense向量：用于语义相似度检索
+    - Sparse向量：用于关键词精确匹配（BM25风格）
+    - ColBERT向量：用于细粒度交互匹配（可选）
+
+    特点：
+    - 支持100+语言（中英双语优化）
+    - 8192 tokens超长上下文
+    - Dense+Sparse双路检索能力
+
+    GitHub: https://github.com/FlagOpen/FlagEmbedding
+    """
+
+    def __init__(
+        self,
+        model_name: str = "BAAI/bge-m3",
+        use_fp16: bool = True,
+        device: Optional[str] = None,
+        batch_size: int = 12,
+        max_length: int = 8192,
+        local_model_path: Optional[str] = None
+    ):
+        """
+        初始化BGE-M3嵌入模型
+
+        Args:
+            model_name: 模型名称（如果使用本地路径，此参数会被忽略）
+            use_fp16: 是否使用FP16加速
+            device: 设备类型（cuda/cpu），默认自动选择
+            batch_size: 批处理大小
+            max_length: 最大序列长度
+            local_model_path: 本地模型路径（可选，优先使用）
+        """
+        self.use_fp16 = use_fp16
+        self.batch_size = batch_size
+        self.max_length = max_length
+
+        # 确定模型路径（优先使用本地路径）
+        if local_model_path and os.path.exists(local_model_path):
+            self.model_path = local_model_path
+            self.model_name = "local"
+            logger.info(f"使用本地模型路径: {local_model_path}")
+        else:
+            # 检查多个可能的本地路径
+            found_local = False
+            for path in LOCAL_MODEL_PATHS:
+                if os.path.exists(path) and os.path.exists(os.path.join(path, "config.json")):
+                    self.model_path = path
+                    self.model_name = "local"
+                    logger.info(f"使用本地模型路径: {path}")
+                    found_local = True
+                    break
+
+            if not found_local:
+                self.model_path = model_name
+                self.model_name = model_name
+                logger.info(f"使用远程模型: {model_name}")
+
+        # 自动选择设备
+        if device is None:
+            self.device = "cuda" if torch.cuda.is_available() else "cpu"
+        else:
+            self.device = device
+
+        logger.info(f"初始化BGE-M3模型, 设备: {self.device}")
+
+        self.model = None
+        self._load_model()
+
+    def _load_model(self):
+        """加载嵌入模型"""
+        try:
+            from FlagEmbedding import BGEM3FlagModel
+
+            self.model = BGEM3FlagModel(
+                self.model_path,
+                use_fp16=self.use_fp16,
+                device=self.device
+            )
+
+            logger.success(f"BGE-M3模型加载成功")
+
+        except ImportError:
+            logger.warning("FlagEmbedding库未安装，请运行: pip install FlagEmbedding")
+            raise
+        except Exception as e:
+            logger.error(f"模型加载失败: {e}")
+            raise
+
+    def embed(
+        self,
+        texts: List[str],
+        return_dense: bool = True,
+        return_sparse: bool = True,
+        return_colbert_vecs: bool = False
+    ) -> EmbeddingResult:
+        """
+        对文本列表生成嵌入向量
+
+        Args:
+            texts: 文本列表
+            return_dense: 是否返回Dense向量
+            return_sparse: 是否返回Sparse向量
+            return_colbert_vecs: 是否返回ColBERT向量
+
+        Returns:
+            EmbeddingResult: 嵌入结果
+        """
+        if not texts:
+            logger.warning("输入文本列表为空")
+            return EmbeddingResult(
+                dense_embeddings=np.array([]),
+                sparse_embeddings=[],
+                texts=[],
+                dim=0
+            )
+
+        logger.info(f"开始嵌入{len(texts)}个文本块")
+
+        try:
+            # 执行嵌入
+            embeddings = self.model.encode(
+                texts,
+                batch_size=self.batch_size,
+                max_length=self.max_length,
+                return_dense=return_dense,
+                return_sparse=return_sparse,
+                return_colbert_vecs=return_colbert_vecs
+            )
+
+            # 提取结果
+            dense_embeddings = embeddings.get('dense_vecs', np.array([]))
+            sparse_embeddings = embeddings.get('lexical_weights', [])
+
+            # 获取维度
+            dim = dense_embeddings.shape[1] if len(dense_embeddings) > 0 else 1024
+
+            logger.success(f"嵌入完成，向量维度: {dim}")
+
+            return EmbeddingResult(
+                dense_embeddings=dense_embeddings,
+                sparse_embeddings=sparse_embeddings,
+                texts=texts,
+                dim=dim
+            )
+
+        except Exception as e:
+            logger.error(f"嵌入失败: {e}")
+            raise
+
+    def embed_single(self, text: str) -> Dict[str, Union[np.ndarray, Dict]]:
+        """
+        对单个文本生成嵌入向量
+
+        Args:
+            text: 输入文本
+
+        Returns:
+            Dict: 包含dense和sparse向量
+        """
+        result = self.embed([text])
+        return {
+            'dense': result.dense_embeddings[0],
+            'sparse': result.sparse_embeddings[0] if result.sparse_embeddings else {},
+            'dim': result.dim
+        }
+
+    def embed_dense(self, texts: List[str]) -> np.ndarray:
+        """只生成Dense向量"""
+        result = self.embed(texts, return_sparse=False, return_colbert_vecs=False)
+        return result.dense_embeddings
+
+    def embed_sparse(self, texts: List[str]) -> List[Dict[int, float]]:
+        """只生成Sparse向量"""
+        result = self.embed(texts, return_dense=False, return_colbert_vecs=False)
+        return result.sparse_embeddings
+
+    def embed_query(self, query: str) -> Dict:
+        """
+        对查询文本生成嵌入（用于检索）
+
+        Args:
+            query: 查询文本
+
+        Returns:
+            Dict: 包含dense和sparse向量
+        """
+        return self.embed_single(query)
+
+    def compute_similarity(
+        self,
+        query_embedding: np.ndarray,
+        doc_embeddings: np.ndarray,
+        metric: str = "cosine"
+    ) -> np.ndarray:
+        """
+        计算查询与文档的相似度
+
+        Args:
+            query_embedding: 查询向量
+            doc_embeddings: 文档向量矩阵
+            metric: 相似度度量（cosine/dot）
+
+        Returns:
+            np.ndarray: 相似度分数数组
+        """
+        if metric == "cosine":
+            # 余弦相似度
+            query_norm = np.linalg.norm(query_embedding)
+            doc_norms = np.linalg.norm(doc_embeddings, axis=1)
+
+            similarities = np.dot(doc_embeddings, query_embedding) / (doc_norms * query_norm)
+
+        elif metric == "dot":
+            # 点积相似度
+            similarities = np.dot(doc_embeddings, query_embedding)
+
+        else:
+            raise ValueError(f"不支持的相似度度量: {metric}")
+
+        return similarities
+
+    def sparse_similarity(
+        self,
+        query_sparse: Dict[int, float],
+        doc_sparse: Dict[int, float]
+    ) -> float:
+        """
+        计算Sparse向量的相似度（BM25风格）
+
+        Args:
+            query_sparse: 查询的Sparse向量（词ID -> 权重）
+            doc_sparse: 文档的Sparse向量
+
+        Returns:
+            float: 相似度分数
+        """
+        # 计算交集词的点积
+        common_keys = set(query_sparse.keys()) & set(doc_sparse.keys())
+
+        score = sum(query_sparse[k] * doc_sparse[k] for k in common_keys)
+        return score
+
+
+def embed_texts(
+    texts: List[str],
+    model_name: str = "BAAI/bge-m3",
+    **kwargs
+) -> EmbeddingResult:
+    """便捷函数：对文本列表生成嵌入"""
+    embedder = BGEM3Embedder(model_name=model_name, **kwargs)
+    return embedder.embed(texts)
+
+
+def embed_single_text(
+    text: str,
+    model_name: str = "BAAI/bge-m3",
+    **kwargs
+) -> Dict:
+    """便捷函数：对单个文本生成嵌入"""
+    embedder = BGEM3Embedder(model_name=model_name, **kwargs)
+    return embedder.embed_single(text)

backend/app/services/embedding/text_chunker.py

@@ -0,0 +1,449 @@
+# src/services/embedding/text_chunker.py
+"""智能分块器 - 章节级+条款级双粒度切割"""
+
+import re
+from typing import List, Dict, Optional, Tuple
+from dataclasses import dataclass, field
+from loguru import logger
+
+
+@dataclass
+class ChunkMetadata:
+    """分块元数据"""
+    doc_id: str = ""
+    doc_name: str = ""
+    chunk_id: str = ""
+    section_number: str = ""  # 章节编号（如 "第一章"）
+    section_title: str = ""   # 章节标题
+    clause_number: str = ""   # 条款编号（如 "第一条"）
+    page_number: int = 0
+    start_position: int = 0   # 在原文中的起始位置
+    end_position: int = 0     # 在原文中的结束位置
+    regulation_type: str = ""  # 法规类型
+    version: str = ""
+
+
+@dataclass
+class TextChunk:
+    """文本分块"""
+    content: str
+    metadata: ChunkMetadata
+    token_count: int = 0  # 估算的token数量
+
+
+class RegulationChunker:
+    """
+    法规文档智能分块器
+
+    实现章节级/条款级双粒度切割，适配国标GB文档结构：
+    - 国标文档通常有明确的层级结构：章 > 节 > 条
+    - 每个条款应作为一个独立的语义单元
+    - 保留条款完整性，避免跨条款截断
+    """
+
+    # 法规标题模式
+    CHAPTER_PATTERN = re.compile(r'^第[一二三四五六七八九十百]+章\s+[^\n]+')
+    SECTION_PATTERN = re.compile(r'^第[一二三四五六七八九十百]+节\s+[^\n]+')
+    CLAUSE_PATTERN = re.compile(r'^第[一二三四五六七八九十百]+条\s')
+
+    # 条款子项模式
+    SUB_ITEM_PATTERN = re.compile(r'^[\(（][一二三四五六七八九十]+[\)）]\s')
+    NUMBER_ITEM_PATTERN = re.compile(r'^[\d]+[\.、]\s')
+
+    def __init__(
+        self,
+        chunk_size: int = 512,
+        chunk_overlap: int = 50,
+        max_chunk_size: int = 2048,
+        min_chunk_size: int = 100
+    ):
+        """
+        初始化分块器
+
+        Args:
+            chunk_size: 默认分块大小（字符数）
+            chunk_overlap: 分块重叠大小
+            max_chunk_size: 最大分块大小（防止单个条款过长）
+            min_chunk_size: 最小分块大小（防止碎片化）
+        """
+        self.chunk_size = chunk_size
+        self.chunk_overlap = chunk_overlap
+        self.max_chunk_size = max_chunk_size
+        self.min_chunk_size = min_chunk_size
+
+    def chunk_document(
+        self,
+        markdown_text: str,
+        doc_id: str = "",
+        doc_name: str = "",
+        regulation_type: str = "",
+        version: str = ""
+    ) -> List[TextChunk]:
+        """
+        对法规文档进行智能分块
+
+        Args:
+            markdown_text: Markdown格式的文档内容
+            doc_id: 文档ID
+            doc_name: 文档名称
+            regulation_type: 法规类型
+            version: 文档版本
+
+        Returns:
+            List[TextChunk]: 分块列表
+        """
+        logger.info(f"开始分块文档: {doc_name}")
+
+        # 1. 按章节分割（一级分块）
+        sections = self._split_by_sections(markdown_text)
+
+        # 2. 在每个章节内按条款分割（二级分块）
+        chunks = []
+        global_position = 0
+
+        for section_num, section_title, section_content, section_start in sections:
+            # 在章节内按条款分割
+            clause_chunks = self._split_by_clauses(
+                section_content,
+                section_num,
+                section_title,
+                section_start + global_position
+            )
+
+            for chunk_content, clause_num, clause_title, start_pos, end_pos in clause_chunks:
+                # 处理过长的条款（进一步细分）
+                if len(chunk_content) > self.max_chunk_size:
+                    sub_chunks = self._split_long_clause(
+                        chunk_content,
+                        clause_num,
+                        clause_title
+                    )
+                    for sub_content, sub_start, sub_end in sub_chunks:
+                        chunk = self._create_chunk(
+                            sub_content,
+                            doc_id,
+                            doc_name,
+                            section_num,
+                            section_title,
+                            clause_num,
+                            sub_start + start_pos,
+                            sub_end + start_pos,
+                            regulation_type,
+                            version
+                        )
+                        chunks.append(chunk)
+                else:
+                    chunk = self._create_chunk(
+                        chunk_content,
+                        doc_id,
+                        doc_name,
+                        section_num,
+                        section_title,
+                        clause_num,
+                        start_pos,
+                        end_pos,
+                        regulation_type,
+                        version
+                    )
+                    chunks.append(chunk)
+
+        logger.success(f"分块完成，共{len(chunks)}个chunk")
+        return chunks
+
+    def _split_by_sections(self, markdown_text: str) -> List[Tuple[str, str, str, int]]:
+        """
+        按章节分割文档
+
+        Returns:
+            List of (section_number, section_title, section_content, start_position)
+        """
+        sections = []
+        lines = markdown_text.split('\n')
+
+        current_section_num = ""
+        current_section_title = ""
+        current_section_content = []
+        current_section_start = 0
+
+        for i, line in enumerate(lines):
+            # 检测章节标题
+            chapter_match = self.CHAPTER_PATTERN.match(line.strip())
+            section_match = self.SECTION_PATTERN.match(line.strip())
+
+            if chapter_match or section_match:
+                # 保存上一个章节
+                if current_section_content:
+                    content = '\n'.join(current_section_content)
+                    sections.append((
+                        current_section_num,
+                        current_section_title,
+                        content,
+                        current_section_start
+                    ))
+
+                # 开始新章节
+                current_section_start = sum(len(l) + 1 for l in lines[:i])
+                current_section_content = []
+
+                if chapter_match:
+                    current_section_num = line.strip()
+                    current_section_title = self._extract_title(line.strip())
+                else:
+                    current_section_num = line.strip()
+                    current_section_title = self._extract_title(line.strip())
+
+            current_section_content.append(line)
+
+        # 保存最后一个章节
+        if current_section_content:
+            content = '\n'.join(current_section_content)
+            sections.append((
+                current_section_num,
+                current_section_title,
+                content,
+                current_section_start
+            ))
+
+        # 如果没有检测到章节，将整个文档作为一个大章节
+        if not sections:
+            sections.append((
+                "",
+                "全文",
+                markdown_text,
+                0
+            ))
+
+        return sections
+
+    def _split_by_clauses(
+        self,
+        section_content: str,
+        section_num: str,
+        section_title: str,
+        section_start: int
+    ) -> List[Tuple[str, str, str, int, int]]:
+        """
+        在章节内按条款分割
+
+        Returns:
+            List of (content, clause_number, clause_title, start_position, end_position)
+        """
+        clauses = []
+        lines = section_content.split('\n')
+
+        current_clause_num = ""
+        current_clause_title = ""
+        current_clause_content = []
+        current_clause_start = section_start
+
+        for i, line in enumerate(lines):
+            # 检测条款标题
+            clause_match = self.CLAUSE_PATTERN.match(line.strip())
+
+            if clause_match:
+                # 保存上一个条款
+                if current_clause_content:
+                    content = '\n'.join(current_clause_content)
+                    end_pos = current_clause_start + len(content)
+                    clauses.append((
+                        content,
+                        current_clause_num,
+                        current_clause_title,
+                        current_clause_start,
+                        end_pos
+                    ))
+
+                # 开始新条款
+                current_clause_start = section_start + sum(len(l) + 1 for l in lines[:i])
+                current_clause_content = []
+                current_clause_num = self._extract_clause_number(line.strip())
+                current_clause_title = line.strip()
+
+            current_clause_content.append(line)
+
+        # 保存最后一个条款
+        if current_clause_content:
+            content = '\n'.join(current_clause_content)
+            end_pos = current_clause_start + len(content)
+            clauses.append((
+                content,
+                current_clause_num,
+                current_clause_title,
+                current_clause_start,
+                end_pos
+            ))
+
+        # 如果没有检测到条款，将整个章节作为一个条款
+        if not clauses:
+            clauses.append((
+                section_content,
+                "",
+                section_title,
+                section_start,
+                section_start + len(section_content)
+            ))
+
+        return clauses
+
+    def _split_long_clause(
+        self,
+        content: str,
+        clause_num: str,
+        clause_title: str
+    ) -> List[Tuple[str, int, int]]:
+        """
+        分割过长的条款内容
+
+        按条款子项或段落分割，保持语义完整性
+        """
+        sub_chunks = []
+        lines = content.split('\n')
+
+        # 检测是否有子项结构
+        has_sub_items = any(
+            self.SUB_ITEM_PATTERN.match(line.strip()) or
+            self.NUMBER_ITEM_PATTERN.match(line.strip())
+            for line in lines
+        )
+
+        if has_sub_items:
+            # 按子项分割
+            current_sub_content = []
+            current_sub_start = 0
+
+            for i, line in enumerate(lines):
+                is_sub_item = (
+                    self.SUB_ITEM_PATTERN.match(line.strip()) or
+                    self.NUMBER_ITEM_PATTERN.match(line.strip())
+                )
+
+                if is_sub_item and current_sub_content:
+                    sub_content = '\n'.join(current_sub_content)
+                    sub_end = current_sub_start + len(sub_content)
+                    if len(sub_content) >= self.min_chunk_size:
+                        sub_chunks.append((sub_content, current_sub_start, sub_end))
+                    current_sub_content = []
+                    current_sub_start = sum(len(l) + 1 for l in lines[:i])
+
+                current_sub_content.append(line)
+
+            # 保存最后一个子项
+            if current_sub_content:
+                sub_content = '\n'.join(current_sub_content)
+                sub_end = current_sub_start + len(sub_content)
+                sub_chunks.append((sub_content, current_sub_start, sub_end))
+
+        else:
+            # 按段落分割（滑动窗口）
+            paragraphs = []
+            current_para = []
+
+            for line in lines:
+                if line.strip():
+                    current_para.append(line)
+                else:
+                    if current_para:
+                        paragraphs.append('\n'.join(current_para))
+                        current_para = []
+
+            if current_para:
+                paragraphs.append('\n'.join(current_para))
+
+            # 合并段落直到达到chunk_size
+            current_chunk = []
+            current_length = 0
+            chunk_start = 0
+
+            for para in paragraphs:
+                if current_length + len(para) > self.chunk_size and current_chunk:
+                    chunk_content = '\n'.join(current_chunk)
+                    chunk_end = chunk_start + len(chunk_content)
+                    sub_chunks.append((chunk_content, chunk_start, chunk_end))
+                    current_chunk = []
+                    current_length = 0
+                    chunk_start = chunk_end
+
+                current_chunk.append(para)
+                current_length += len(para)
+
+            # 保存最后一个chunk
+            if current_chunk:
+                chunk_content = '\n'.join(current_chunk)
+                chunk_end = chunk_start + len(chunk_content)
+                sub_chunks.append((chunk_content, chunk_start, chunk_end))
+
+        return sub_chunks
+
+    def _extract_title(self, header_line: str) -> str:
+        """从标题行提取标题内容"""
+        # 移除"第X章"、"第X节"前缀
+        title = re.sub(r'^第[一二三四五六七八九十百]+[章节]\s+', '', header_line)
+        return title.strip()
+
+    def _extract_clause_number(self, clause_line: str) -> str:
+        """从条款行提取条款编号"""
+        match = self.CLAUSE_PATTERN.match(clause_line)
+        if match:
+            return match.group(0).strip()
+        return ""
+
+    def _create_chunk(
+        self,
+        content: str,
+        doc_id: str,
+        doc_name: str,
+        section_num: str,
+        section_title: str,
+        clause_num: str,
+        start_pos: int,
+        end_pos: int,
+        regulation_type: str,
+        version: str
+    ) -> TextChunk:
+        """创建文本分块"""
+        # 清理内容
+        content = content.strip()
+
+        # 计算估算token数（中文约1.5字符/token）
+        token_count = int(len(content) * 0.7)  # 简化估算
+
+        # 生成chunk_id
+        chunk_id = f"{doc_id}_{section_num}_{clause_num}_{start_pos}"
+
+        metadata = ChunkMetadata(
+            doc_id=doc_id,
+            doc_name=doc_name,
+            chunk_id=chunk_id,
+            section_number=section_num,
+            section_title=section_title,
+            clause_number=clause_num,
+            start_position=start_pos,
+            end_position=end_pos,
+            regulation_type=regulation_type,
+            version=version
+        )
+
+        return TextChunk(
+            content=content,
+            metadata=metadata,
+            token_count=token_count
+        )
+
+
+def chunk_regulation_document(
+    markdown_text: str,
+    doc_id: str = "",
+    doc_name: str = "",
+    regulation_type: str = "",
+    version: str = "",
+    chunk_size: int = 512
+) -> List[TextChunk]:
+    """便捷函数：对法规文档进行分块"""
+    chunker = RegulationChunker(chunk_size=chunk_size)
+    return chunker.chunk_document(
+        markdown_text,
+        doc_id,
+        doc_name,
+        regulation_type,
+        version
+    )

backend/app/services/llm/__init__.py

@@ -0,0 +1,15 @@
+# src/services/llm/__init__.py
+"""LLM服务模块"""
+
+from .llm_factory import LLMFactory, get_llm_client
+from .base_client import BaseLLMClient, LLMResponse, LLMConfig, LLMProvider
+from .deepseek_client import DeepSeekClient
+from .qwen_client import QwenClient, QwenVLClient
+from .document_summarizer import DocumentSummarizer, summarize_document, DocumentSummary
+
+__all__ = [
+    "LLMFactory", "get_llm_client",
+    "BaseLLMClient", "LLMResponse", "LLMConfig", "LLMProvider",
+    "DeepSeekClient", "QwenClient", "QwenVLClient",
+    "DocumentSummarizer", "summarize_document", "DocumentSummary"
+]

backend/app/services/llm/base_client.py

@@ -0,0 +1,116 @@
+# src/services/llm/base_client.py
+"""LLM客户端基类 - 统一接口定义"""
+
+from abc import ABC, abstractmethod
+from dataclasses import dataclass, field
+from typing import List, Dict, Optional, Any
+from enum import Enum
+
+
+class LLMProvider(Enum):
+    """LLM提供商"""
+    DEEPSEEK = "deepseek"
+    QWEN = "qwen"
+    QWEN_VL = "qwen_vl"
+
+
+@dataclass
+class LLMResponse:
+    """LLM响应结果"""
+    content: str
+    model: str
+    usage: Dict[str, int] = field(default_factory=dict)
+    finish_reason: str = "stop"
+    latency_ms: int = 0
+    error: Optional[str] = None
+
+    @property
+    def is_success(self) -> bool:
+        return self.error is None
+
+
+@dataclass
+class LLMConfig:
+    """LLM配置"""
+    provider: LLMProvider
+    model: str
+    api_key: str
+    base_url: str
+    max_tokens: int = 4096
+    temperature: float = 0.7
+    top_p: float = 0.9
+    timeout: int = 300  # 默认超时300秒（摘要/Skills生成可能需要较长时间）
+
+
+class BaseLLMClient(ABC):
+    """LLM客户端基类"""
+
+    def __init__(self, config: LLMConfig):
+        self.config = config
+        self._client = None
+
+    @abstractmethod
+    def _init_client(self):
+        """初始化客户端"""
+        pass
+
+    @abstractmethod
+    def chat(
+        self,
+        messages: List[Dict[str, str]],
+        max_tokens: Optional[int] = None,
+        temperature: Optional[float] = None,
+        **kwargs
+    ) -> LLMResponse:
+        """
+        对话补全
+
+        Args:
+            messages: 对话消息列表 [{"role": "user/assistant/system", "content": "..."}]
+            max_tokens: 最大输出token数
+            temperature: 温度参数
+            **kwargs: 其他参数
+
+        Returns:
+            LLMResponse: 响应结果
+        """
+        pass
+
+    def complete(
+        self,
+        prompt: str,
+        system_prompt: Optional[str] = None,
+        max_tokens: Optional[int] = None,
+        temperature: Optional[float] = None,
+        **kwargs
+    ) -> LLMResponse:
+        """
+        单轮补全（便捷方法）
+
+        Args:
+            prompt: 用户输入
+            system_prompt: 系统提示词
+            max_tokens: 最大输出token数
+            temperature: 温度参数
+
+        Returns:
+            LLMResponse: 响应结果
+        """
+        messages = []
+        if system_prompt:
+            messages.append({"role": "system", "content": system_prompt})
+        messages.append({"role": "user", "content": prompt})
+
+        return self.chat(messages, max_tokens, temperature, **kwargs)
+
+    @abstractmethod
+    def get_available_models(self) -> List[str]:
+        """获取可用模型列表"""
+        pass
+
+    def estimate_tokens(self, text: str) -> int:
+        """估算文本token数（粗略估计）"""
+        # 中文字符约1.5 token，英文约0.25 token
+        chinese_chars = sum(1 for c in text if '一' <= c <= '鿿')
+        other_chars = len(text) - chinese_chars
+        return int(chinese_chars * 1.5 + other_chars * 0.25)

backend/app/services/llm/deepseek_client.py

@@ -0,0 +1,130 @@
+# src/services/llm/deepseek_client.py
+"""DeepSeek LLM客户端 - OpenAI兼容API"""
+
+import time
+from typing import List, Dict, Optional
+from loguru import logger
+import httpx
+
+from .base_client import BaseLLMClient, LLMResponse, LLMConfig, LLMProvider
+
+
+class DeepSeekClient(BaseLLMClient):
+    """
+    DeepSeek API客户端（OpenAI兼容格式）
+
+    支持模型：
+    - deepseek-chat
+    - deepseek-coder
+    - deepseek-reasoner
+    - deepseek-v3
+    - deepseek-v3.2
+    - deepseek-v4-flash
+    """
+
+    SUPPORTED_MODELS = [
+        "deepseek-chat",
+        "deepseek-coder",
+        "deepseek-reasoner",
+        "deepseek-v3",
+        "deepseek-v3.2",
+        "deepseek-v4-flash"
+    ]
+
+    def __init__(self, config: LLMConfig):
+        if config.provider != LLMProvider.DEEPSEEK:
+            raise ValueError(f"配置provider应为DEEPSEEK，实际为{config.provider}")
+        super().__init__(config)
+        self._init_client()
+
+    def _init_client(self):
+        """初始化HTTP客户端"""
+        self._client = httpx.Client(
+            base_url=self.config.base_url,
+            headers={
+                "Authorization": f"Bearer {self.config.api_key}",
+                "Content-Type": "application/json"
+            },
+            timeout=self.config.timeout
+        )
+        logger.info(f"DeepSeek客户端初始化完成: {self.config.base_url} - {self.config.model}")
+
+    def chat(
+        self,
+        messages: List[Dict[str, str]],
+        max_tokens: Optional[int] = None,
+        temperature: Optional[float] = None,
+        **kwargs
+    ) -> LLMResponse:
+        """对话补全"""
+        start_time = time.time()
+
+        try:
+            payload = {
+                "model": self.config.model,
+                "messages": messages,
+                "max_tokens": max_tokens or self.config.max_tokens,
+                "temperature": temperature or self.config.temperature,
+                "top_p": kwargs.get("top_p", self.config.top_p),
+                "stream": False
+            }
+
+            response = self._client.post("/chat/completions", json=payload)
+            response.raise_for_status()
+
+            data = response.json()
+
+            latency_ms = int((time.time() - start_time) * 1000)
+
+            choices = data.get("choices", [{}])
+            message = choices[0].get("message", {})
+
+            return LLMResponse(
+                content=message.get("content", ""),
+                model=data.get("model", self.config.model),
+                usage=data.get("usage", {}),
+                finish_reason=choices[0].get("finish_reason", "stop"),
+                latency_ms=latency_ms
+            )
+
+        except httpx.HTTPStatusError as e:
+            logger.error(f"DeepSeek API错误: {e.response.status_code} - {e.response.text}")
+            return LLMResponse(
+                content="",
+                model=self.config.model,
+                error=f"API错误: {e.response.status_code} - {e.response.text[:200]}"
+            )
+
+        except Exception as e:
+            logger.error(f"DeepSeek调用失败: {e}")
+            return LLMResponse(
+                content="",
+                model=self.config.model,
+                error=str(e)
+            )
+
+    def get_available_models(self) -> List[str]:
+        """获取可用模型列表"""
+        return self.SUPPORTED_MODELS
+
+    def close(self):
+        """关闭客户端"""
+        if self._client:
+            self._client.close()
+
+
+def create_deepseek_client(
+    api_key: str,
+    model: str = "deepseek-v4-flash",
+    base_url: str = "http://6.86.80.4:30080/v1",
+    **kwargs
+) -> DeepSeekClient:
+    """便捷函数：创建DeepSeek客户端"""
+    config = LLMConfig(
+        provider=LLMProvider.DEEPSEEK,
+        model=model,
+        api_key=api_key,
+        base_url=base_url,
+        **kwargs
+    )
+    return DeepSeekClient(config)

backend/app/services/llm/document_summarizer.py

@@ -0,0 +1,231 @@
+# src/services/llm/document_summarizer.py
+"""文档摘要生成服务 - LLM生成法规文档摘要"""
+
+from typing import Dict, Optional
+from dataclasses import dataclass
+from loguru import logger
+
+from app.services.llm import get_llm_client, BaseLLMClient
+from app.services.rag.prompt_templates import get_prompt_template
+from app.config.settings import settings
+
+
+@dataclass
+class DocumentSummary:
+    """文档摘要结果"""
+    doc_name: str
+    summary: str
+    applicable_scope: str
+    key_clauses: list
+    key_terms: list
+    compliance_points: list
+    model: str
+    latency_ms: int
+    error: Optional[str] = None
+
+    @property
+    def is_success(self) -> bool:
+        return self.error is None
+
+
+class DocumentSummarizer:
+    """
+    文档摘要生成器
+
+    功能：
+    - 生成法规文档的核心要点摘要
+    - 提取适用范围
+    - 突出关键条款
+    - 列出合规要点
+
+    使用示例：
+        summarizer = DocumentSummarizer()
+        result = summarizer.summarize("GB 7258-2017", markdown_content)
+        print(result.summary)
+    """
+
+    def __init__(
+        self,
+        provider: str = None,
+        model: str = None,
+        max_tokens: int = None
+    ):
+        """
+        初始化摘要生成器
+
+        Args:
+            provider: LLM提供商
+            model: LLM模型名称
+            max_tokens: 最大输出token数
+        """
+        self.provider = provider or settings.llm_provider
+        self.model = model or settings.llm_model
+        self.max_tokens = max_tokens or settings.rag_summary_max_tokens
+
+        # LLM客户端（延迟加载）
+        self.llm: Optional[BaseLLMClient] = None
+
+        logger.info(f"摘要生成器初始化: provider={self.provider}, model={self.model}")
+
+    def _init_llm(self):
+        """延迟初始化LLM"""
+        if self.llm is None:
+            self.llm = get_llm_client(
+                provider=self.provider,
+                model=self.model
+            )
+
+    def summarize(
+        self,
+        doc_name: str,
+        content: str,
+        regulation_type: str = "",
+        max_tokens: Optional[int] = None
+    ) -> DocumentSummary:
+        """
+        生成文档摘要
+
+        Args:
+            doc_name: 文档名称
+            content: 文档内容（Markdown格式）
+            regulation_type: 法规类型
+            max_tokens: 最大输出token数
+
+        Returns:
+            DocumentSummary: 摘要结果
+        """
+        import time
+        start_time = time.time()
+
+        logger.info(f"生成文档摘要: {doc_name}")
+
+        try:
+            self._init_llm()
+
+            # 使用摘要模板
+            template = get_prompt_template("document_summary")
+
+            # 构建用户消息
+            user_content = template.user_template.format(
+                doc_name=doc_name,
+                content=content[:8000]  # 截取前8000字符（避免超出token限制）
+            )
+
+            # 调用LLM
+            response = self.llm.chat(
+                messages=[
+                    {"role": "system", "content": template.system_prompt},
+                    {"role": "user", "content": user_content}
+                ],
+                max_tokens=max_tokens or self.max_tokens,
+                temperature=0.3  # 低温度保证摘要准确性
+            )
+
+            latency_ms = int((time.time() - start_time) * 1000)
+
+            if not response.is_success:
+                return DocumentSummary(
+                    doc_name=doc_name,
+                    summary="",
+                    applicable_scope="",
+                    key_clauses=[],
+                    key_terms=[],
+                    compliance_points=[],
+                    model=self.model,
+                    latency_ms=latency_ms,
+                    error=response.error
+                )
+
+            # 解析摘要结构
+            summary_data = self._parse_summary(response.content)
+
+            logger.success(f"摘要生成完成: {doc_name}, {latency_ms}ms")
+
+            return DocumentSummary(
+                doc_name=doc_name,
+                summary=summary_data.get("summary", response.content),
+                applicable_scope=summary_data.get("applicable_scope", ""),
+                key_clauses=summary_data.get("key_clauses", []),
+                key_terms=summary_data.get("key_terms", []),
+                compliance_points=summary_data.get("compliance_points", []),
+                model=response.model,
+                latency_ms=latency_ms
+            )
+
+        except Exception as e:
+            logger.error(f"摘要生成失败: {e}")
+            return DocumentSummary(
+                doc_name=doc_name,
+                summary="",
+                applicable_scope="",
+                key_clauses=[],
+                key_terms=[],
+                compliance_points=[],
+                model=self.model,
+                latency_ms=0,
+                error=str(e)
+            )
+
+    def _parse_summary(self, content: str) -> Dict:
+        """解析摘要内容（提取结构化信息）"""
+        result = {
+            "summary": content,
+            "applicable_scope": "",
+            "key_clauses": [],
+            "key_terms": [],
+            "compliance_points": []
+        }
+
+        # 简单解析（提取关键信息）
+        lines = content.split("\n")
+
+        for line in lines:
+            line = line.strip()
+
+            # 提取适用范围
+            if "适用范围" in line or "适用对象" in line:
+                result["applicable_scope"] = line.split("：")[-1].strip() if "：" in line else line.split(":")[-1].strip()
+
+            # 提取关键条款
+            if line.startswith("- 【条款") or line.startswith("【条款"):
+                result["key_clauses"].append(line)
+
+            # 提取关键术语
+            if "关键术语" in line or "术语定义" in line:
+                # 继续读取后续几行
+                pass
+
+            # 提取合规要点
+            if "合规要点" in line or "必须满足" in line:
+                pass
+
+        return result
+
+    def batch_summarize(
+        self,
+        documents: list
+    ) -> list:
+        """
+        批量生成摘要
+
+        Args:
+            documents: 文档列表 [{"doc_name": str, "content": str}, ...]
+
+        Returns:
+            list: 摘要结果列表
+        """
+        results = []
+        for doc in documents:
+            result = self.summarize(doc["doc_name"], doc["content"])
+            results.append(result)
+        return results
+
+
+def summarize_document(
+    doc_name: str,
+    content: str,
+    **kwargs
+) -> DocumentSummary:
+    """便捷函数：生成文档摘要"""
+    summarizer = DocumentSummarizer(**kwargs)
+    return summarizer.summarize(doc_name, content)

backend/app/services/llm/llm_factory.py

@@ -0,0 +1,258 @@
+# src/services/llm/llm_factory.py
+"""LLM工厂 - 统一创建和管理LLM客户端"""
+
+from typing import Optional, Dict, Any
+from loguru import logger
+from functools import lru_cache
+
+from .base_client import BaseLLMClient, LLMConfig, LLMProvider, LLMResponse
+from .deepseek_client import DeepSeekClient
+from .qwen_client import QwenClient, QwenVLClient
+
+
+# 默认模型映射
+DEFAULT_MODELS = {
+    LLMProvider.DEEPSEEK: "deepseek-v4-flash",
+    LLMProvider.QWEN: "qwen3.5-flash",
+    LLMProvider.QWEN_VL: "qwen3-vl-plus"
+}
+
+# API基础URL（使用统一代理服务）
+DEFAULT_BASE_URLS = {
+    LLMProvider.DEEPSEEK: "http://6.86.80.4:30080/v1",
+    LLMProvider.QWEN: "http://6.86.80.4:30080/v1",
+    LLMProvider.QWEN_VL: "http://6.86.80.4:30080/v1"
+}
+
+
+class LLMFactory:
+    """
+    LLM客户端工厂（支持全局缓存）
+
+    支持的提供商和模型：
+    - DeepSeek: deepseek-chat (DeepSeek-V3), deepseek-coder
+    - Qwen: qwen-turbo, qwen-plus, qwen-max, qwen-long
+    - QwenVL: qwen-vl-plus, qwen-vl-max (多模态)
+
+    使用示例：
+        factory = LLMFactory()
+
+        # 使用默认配置
+        client = factory.create("deepseek")
+
+        # 自定义配置
+        client = factory.create("qwen", model="qwen-max", temperature=0.5)
+
+        # 调用LLM
+        response = client.complete("你好，介绍一下自己")
+    """
+
+    # 全局客户端缓存（类级别，跨实例共享）
+    _global_instances: Dict[str, BaseLLMClient] = {}
+
+    def __init__(self):
+        self._config_cache: Dict[str, Any] = {}
+
+    def create(
+        self,
+        provider: str,
+        api_key: Optional[str] = None,
+        model: Optional[str] = None,
+        base_url: Optional[str] = None,
+        max_tokens: int = 4096,
+        temperature: float = 0.7,
+        **kwargs
+    ) -> BaseLLMClient:
+        """
+        创建LLM客户端
+
+        Args:
+            provider: 提供商名称 ("deepseek", "qwen", "qwen_vl")
+            api_key: API密钥（如未提供，从环境变量获取）
+            model: 模型名称（如未提供，使用默认模型）
+            base_url: API基础URL
+            max_tokens: 最大输出token数
+            temperature: 温度参数
+            **kwargs: 其他配置参数
+
+        Returns:
+            BaseLLMClient: LLM客户端实例
+        """
+        provider_enum = self._parse_provider(provider)
+
+        # 获取配置
+        api_key = api_key or self._get_api_key(provider_enum)
+        model = model or DEFAULT_MODELS.get(provider_enum)
+        base_url = base_url or DEFAULT_BASE_URLS.get(provider_enum)
+
+        if not api_key:
+            raise ValueError(f"缺少API密钥，请设置环境变量或传入api_key参数")
+
+        # 检查全局缓存
+        cache_key = f"{provider}_{model}"
+        if cache_key in LLMFactory._global_instances:
+            logger.debug(f"使用缓存的LLM客户端: {cache_key}")
+            return LLMFactory._global_instances[cache_key]
+
+        config = LLMConfig(
+            provider=provider_enum,
+            model=model,
+            api_key=api_key,
+            base_url=base_url,
+            max_tokens=max_tokens,
+            temperature=temperature,
+            **kwargs
+        )
+
+        # 创建客户端
+        client = self._create_client(config)
+
+        # 缓存到全局实例
+        LLMFactory._global_instances[cache_key] = client
+
+        logger.info(f"LLM客户端创建成功并缓存: {provider} - {model}")
+        return client
+
+    def _parse_provider(self, provider: str) -> LLMProvider:
+        """解析提供商名称"""
+        provider_map = {
+            "deepseek": LLMProvider.DEEPSEEK,
+            "deepseek-v3": LLMProvider.DEEPSEEK,
+            "deepseek_chat": LLMProvider.DEEPSEEK,
+            "qwen": LLMProvider.QWEN,
+            "qwen-turbo": LLMProvider.QWEN,
+            "qwen-plus": LLMProvider.QWEN,
+            "qwen-max": LLMProvider.QWEN,
+            "qwen3.5-flash": LLMProvider.QWEN,
+            "qwen3.5-plus": LLMProvider.QWEN,
+            "qwen_vl": LLMProvider.QWEN_VL,
+            "qwen-vl": LLMProvider.QWEN_VL,
+            "qwen-vl-plus": LLMProvider.QWEN_VL,
+            "qwen-vl-max": LLMProvider.QWEN_VL
+        }
+
+        provider_lower = provider.lower()
+        if provider_lower not in provider_map:
+            raise ValueError(f"不支持的提供商: {provider}，支持的: {list(provider_map.keys())}")
+
+        return provider_map[provider_lower]
+
+    def _get_api_key(self, provider: LLMProvider) -> Optional[str]:
+        """从环境变量获取API密钥"""
+        import os
+
+        key_map = {
+            LLMProvider.DEEPSEEK: ["DEEPSEEK_API_KEY", "OPENAI_API_KEY"],
+            LLMProvider.QWEN: ["QWEN_API_KEY", "DASHSCOPE_API_KEY", "ALIBABA_API_KEY"],
+            LLMProvider.QWEN_VL: ["QWEN_API_KEY", "DASHSCOPE_API_KEY", "ALIBABA_API_KEY"]
+        }
+
+        for key_name in key_map.get(provider, []):
+            api_key = os.getenv(key_name)
+            if api_key:
+                return api_key
+
+        return None
+
+    def _create_client(self, config: LLMConfig) -> BaseLLMClient:
+        """创建具体客户端"""
+        client_map = {
+            LLMProvider.DEEPSEEK: DeepSeekClient,
+            LLMProvider.QWEN: QwenClient,
+            LLMProvider.QWEN_VL: QwenVLClient
+        }
+
+        client_class = client_map.get(config.provider)
+        if not client_class:
+            raise ValueError(f"不支持的提供商: {config.provider}")
+
+        return client_class(config)
+
+    def get_cached(self, provider: str, model: Optional[str] = None) -> Optional[BaseLLMClient]:
+        """获取缓存的客户端"""
+        provider_enum = self._parse_provider(provider)
+        model = model or DEFAULT_MODELS.get(provider_enum)
+        cache_key = f"{provider}_{model}"
+        return LLMFactory._global_instances.get(cache_key)
+
+    def list_available_providers(self) -> Dict[str, list]:
+        """列出可用的提供商和模型"""
+        return {
+            "deepseek": DeepSeekClient.SUPPORTED_MODELS,
+            "qwen": QwenClient.SUPPORTED_MODELS,
+            "qwen_vl": QwenVLClient.SUPPORTED_MODELS
+        }
+
+    @classmethod
+    def preload_clients(cls, providers: list = None):
+        """
+        预加载LLM客户端（应用启动时调用）
+
+        Args:
+            providers: 要预加载的提供商列表，默认加载qwen和deepseek
+        """
+        if providers is None:
+            providers = ["qwen", "deepseek"]
+
+        factory = cls()
+        for provider in providers:
+            try:
+                client = factory.create(provider)
+                logger.success(f"预加载LLM客户端成功: {provider}")
+            except Exception as e:
+                logger.warning(f"预加载LLM客户端失败: {provider} - {e}")
+
+    @classmethod
+    def get_global_client(cls, provider: str, model: Optional[str] = None) -> Optional[BaseLLMClient]:
+        """获取全局缓存的客户端"""
+        provider_lower = provider.lower()
+        # 处理模型名作为provider的情况（如 qwen3.5-flash）
+        if provider_lower.startswith("qwen"):
+            provider_lower = "qwen"
+        model = model or DEFAULT_MODELS.get(LLMProvider.QWEN if provider_lower == "qwen" else LLMProvider.DEEPSEEK)
+        cache_key = f"{provider_lower}_{model}"
+        return cls._global_instances.get(cache_key)
+
+    @classmethod
+    def cleanup(cls):
+        """清理所有缓存的客户端"""
+        for cache_key, client in cls._global_instances.items():
+            try:
+                client.close()
+                logger.debug(f"关闭LLM客户端: {cache_key}")
+            except Exception as e:
+                logger.warning(f"关闭LLM客户端失败: {cache_key} - {e}")
+        cls._global_instances.clear()
+        logger.info("所有LLM客户端已清理")
+
+
+@lru_cache
+def get_llm_factory() -> LLMFactory:
+    """获取LLM工厂实例（缓存）"""
+    return LLMFactory()
+
+
+def get_llm_client(
+    provider: str = "qwen",
+    model: Optional[str] = None,
+    **kwargs
+) -> BaseLLMClient:
+    """
+    便捷函数：获取LLM客户端（优先使用缓存）
+
+    Args:
+        provider: 提供商名称
+        model: 模型名称
+        **kwargs: 其他配置
+
+    Returns:
+        BaseLLMClient: LLM客户端实例
+    """
+    factory = get_llm_factory()
+
+    # 先尝试获取缓存的实例
+    cached = factory.get_cached(provider, model)
+    if cached:
+        return cached
+
+    return factory.create(provider, model=model, **kwargs)

backend/app/services/llm/qwen_client.py

@@ -0,0 +1,392 @@
+# src/services/llm/qwen_client.py
+"""Qwen LLM客户端 - 支持OpenAI兼容API格式"""
+
+import time
+import json
+from typing import List, Dict, Optional, Generator, AsyncGenerator
+from loguru import logger
+import httpx
+
+from .base_client import BaseLLMClient, LLMResponse, LLMConfig, LLMProvider
+
+
+class QwenClient(BaseLLMClient):
+    """
+    Qwen API客户端（OpenAI兼容格式）
+
+    支持通过new-api等代理服务调用：
+    - qwen-turbo
+    - qwen-plus
+    - qwen-max
+    - qwen3.5-flash (推荐：快速响应)
+    - qwen3.5-plus
+    - qwen-long
+    - qwen2.5系列
+    """
+
+    SUPPORTED_MODELS = [
+        "qwen-turbo",
+        "qwen-plus",
+        "qwen-max",
+        "qwen-max-longcontext",
+        "qwen-long",
+        "qwen3.5-flash",
+        "qwen3.5-plus",
+        "qwen3-plus",
+        "qwen2.5-72b-instruct",
+        "qwen2.5-32b-instruct",
+        "qwen2.5-14b-instruct",
+        "qwen2.5-7b-instruct"
+    ]
+
+    def __init__(self, config: LLMConfig):
+        if config.provider not in [LLMProvider.QWEN, LLMProvider.QWEN_VL]:
+            raise ValueError(f"配置provider应为Qwen，实际为{config.provider}")
+        super().__init__(config)
+        self._init_client()
+
+    def _init_client(self):
+        """初始化HTTP客户端"""
+        # OpenAI兼容API格式
+        self._client = httpx.Client(
+            base_url=self.config.base_url,
+            headers={
+                "Authorization": f"Bearer {self.config.api_key}",
+                "Content-Type": "application/json"
+            },
+            timeout=self.config.timeout
+        )
+        logger.info(f"Qwen客户端初始化完成: {self.config.base_url} - {self.config.model}")
+
+    def chat(
+        self,
+        messages: List[Dict[str, str]],
+        max_tokens: Optional[int] = None,
+        temperature: Optional[float] = None,
+        **kwargs
+    ) -> LLMResponse:
+        """对话补全（OpenAI兼容格式）"""
+        start_time = time.time()
+
+        try:
+            # OpenAI兼容格式的请求体
+            payload = {
+                "model": self.config.model,
+                "messages": messages,
+                "max_tokens": max_tokens or self.config.max_tokens,
+                "temperature": temperature or self.config.temperature,
+                "top_p": kwargs.get("top_p", self.config.top_p),
+                "stream": False
+            }
+
+            # OpenAI兼容接口路径
+            response = self._client.post("/chat/completions", json=payload)
+            response.raise_for_status()
+
+            data = response.json()
+
+            latency_ms = int((time.time() - start_time) * 1000)
+
+            # OpenAI兼容格式的响应解析
+            choices = data.get("choices", [{}])
+            message = choices[0].get("message", {})
+
+            return LLMResponse(
+                content=message.get("content", ""),
+                model=data.get("model", self.config.model),
+                usage=data.get("usage", {}),
+                finish_reason=choices[0].get("finish_reason", "stop"),
+                latency_ms=latency_ms
+            )
+
+        except httpx.HTTPStatusError as e:
+            logger.error(f"Qwen API错误: {e.response.status_code} - {e.response.text}")
+            return LLMResponse(
+                content="",
+                model=self.config.model,
+                error=f"API错误: {e.response.status_code} - {e.response.text[:200]}"
+            )
+
+        except Exception as e:
+            logger.error(f"Qwen调用失败: {e}")
+            return LLMResponse(
+                content="",
+                model=self.config.model,
+                error=str(e)
+            )
+
+    def stream_chat(
+        self,
+        messages: List[Dict[str, str]],
+        max_tokens: Optional[int] = None,
+        temperature: Optional[float] = None,
+        **kwargs
+    ) -> Generator[str, None, None]:
+        """
+        流式对话补全（SSE格式）
+
+        Yields:
+            str: 每次返回一个文本片段
+
+        使用示例:
+            for chunk in client.stream_chat(messages):
+                print(chunk, end="", flush=True)
+        """
+        try:
+            # OpenAI兼容格式的请求体，启用流式输出
+            payload = {
+                "model": self.config.model,
+                "messages": messages,
+                "max_tokens": max_tokens or self.config.max_tokens,
+                "temperature": temperature or self.config.temperature,
+                "top_p": kwargs.get("top_p", self.config.top_p),
+                "stream": True  # 启用流式输出
+            }
+
+            # 使用stream模式发送请求
+            with self._client.stream("POST", "/chat/completions", json=payload) as response:
+                for line in response.iter_lines():
+                    if line:
+                        line = line.strip()
+                        # SSE格式: data: {...}
+                        if line.startswith("data: "):
+                            data_str = line[6:]  # 移除 "data: " 前缀
+                            if data_str == "[DONE]":
+                                break
+                            try:
+                                data = json.loads(data_str)
+                                choices = data.get("choices", [])
+                                if not choices:
+                                    continue  # 跳过空的choices
+                                delta = choices[0].get("delta", {})
+                                content = delta.get("content", "")
+                                if content:
+                                    yield content
+                            except json.JSONDecodeError:
+                                continue
+
+        except httpx.HTTPStatusError as e:
+            logger.error(f"Qwen流式API错误: {e.response.status_code}")
+            yield f"[ERROR: API返回错误 {e.response.status_code}]"
+
+        except Exception as e:
+            logger.error(f"Qwen流式调用失败: {e}")
+            yield f"[ERROR: {str(e)}]"
+
+    async def async_stream_chat(
+        self,
+        messages: List[Dict[str, str]],
+        max_tokens: Optional[int] = None,
+        temperature: Optional[float] = None,
+        **kwargs
+    ) -> AsyncGenerator[str, None]:
+        """
+        异步流式对话补全（用于FastAPI SSE响应）
+
+        Yields:
+            str: 每次返回一个文本片段
+        """
+        import asyncio
+
+        # 使用同步流式方法，包装为异步
+        for chunk in self.stream_chat(messages, max_tokens, temperature, **kwargs):
+            yield chunk
+            # 给async循环一个小延迟，让其他任务有机会执行
+            await asyncio.sleep(0)
+
+    def get_available_models(self) -> List[str]:
+        """获取可用模型列表"""
+        return self.SUPPORTED_MODELS
+
+    def close(self):
+        """关闭客户端"""
+        if self._client:
+            self._client.close()
+
+
+class QwenVLClient(BaseLLMClient):
+    """
+    Qwen VL多模态客户端（OpenAI兼容格式）
+
+    支持模型：
+    - qwen-vl-plus
+    - qwen-vl-max
+    - qwen3-vl-plus
+    - qwen2-vl-7b-instruct
+    - qwen2-vl-72b-instruct
+    """
+
+    SUPPORTED_MODELS = [
+        "qwen-vl-plus",
+        "qwen-vl-max",
+        "qwen3-vl-plus",
+        "qwen2-vl-7b-instruct",
+        "qwen2-vl-72b-instruct"
+    ]
+
+    def __init__(self, config: LLMConfig):
+        if config.provider != LLMProvider.QWEN_VL:
+            raise ValueError(f"配置provider应为QWEN_VL，实际为{config.provider}")
+        super().__init__(config)
+        self._init_client()
+
+    def _init_client(self):
+        """初始化HTTP客户端"""
+        self._client = httpx.Client(
+            base_url=self.config.base_url,
+            headers={
+                "Authorization": f"Bearer {self.config.api_key}",
+                "Content-Type": "application/json"
+            },
+            timeout=self.config.timeout
+        )
+        logger.info(f"QwenVL客户端初始化完成: {self.config.base_url} - {self.config.model}")
+
+    def chat(
+        self,
+        messages: List[Dict[str, str]],
+        max_tokens: Optional[int] = None,
+        temperature: Optional[float] = None,
+        **kwargs
+    ) -> LLMResponse:
+        """多模态对话补全（OpenAI兼容格式）
+
+        支持图片输入，消息格式：
+        {
+            "role": "user",
+            "content": [
+                {"type": "image_url", "image_url": {"url": "https://example.com/image.jpg"}},
+                {"type": "text", "text": "描述这张图片"}
+            ]
+        }
+        """
+        start_time = time.time()
+
+        try:
+            # OpenAI兼容格式的请求体
+            payload = {
+                "model": self.config.model,
+                "messages": messages,
+                "max_tokens": max_tokens or self.config.max_tokens,
+                "temperature": temperature or self.config.temperature,
+                "top_p": kwargs.get("top_p", self.config.top_p),
+                "stream": False
+            }
+
+            response = self._client.post("/chat/completions", json=payload)
+            response.raise_for_status()
+
+            data = response.json()
+            latency_ms = int((time.time() - start_time) * 1000)
+
+            choices = data.get("choices", [{}])
+            message = choices[0].get("message", {})
+
+            return LLMResponse(
+                content=message.get("content", ""),
+                model=data.get("model", self.config.model),
+                usage=data.get("usage", {}),
+                finish_reason=choices[0].get("finish_reason", "stop"),
+                latency_ms=latency_ms
+            )
+
+        except httpx.HTTPStatusError as e:
+            logger.error(f"QwenVL API错误: {e.response.status_code} - {e.response.text}")
+            return LLMResponse(
+                content="",
+                model=self.config.model,
+                error=f"API错误: {e.response.status_code} - {e.response.text[:200]}"
+            )
+
+        except Exception as e:
+            logger.error(f"QwenVL调用失败: {e}")
+            return LLMResponse(
+                content="",
+                model=self.config.model,
+                error=str(e)
+            )
+
+    def stream_chat(
+        self,
+        messages: List[Dict[str, str]],
+        max_tokens: Optional[int] = None,
+        temperature: Optional[float] = None,
+        **kwargs
+    ) -> Generator[str, None, None]:
+        """流式多模态对话补全"""
+        try:
+            payload = {
+                "model": self.config.model,
+                "messages": messages,
+                "max_tokens": max_tokens or self.config.max_tokens,
+                "temperature": temperature or self.config.temperature,
+                "top_p": kwargs.get("top_p", self.config.top_p),
+                "stream": True
+            }
+
+            with self._client.stream("POST", "/chat/completions", json=payload) as response:
+                for line in response.iter_lines():
+                    if line:
+                        line = line.strip()
+                        if line.startswith("data: "):
+                            data_str = line[6:]
+                            if data_str == "[DONE]":
+                                break
+                            try:
+                                data = json.loads(data_str)
+                                choices = data.get("choices", [])
+                                if not choices:
+                                    continue  # 跳过空的choices
+                                delta = choices[0].get("delta", {})
+                                content = delta.get("content", "")
+                                if content:
+                                    yield content
+                            except json.JSONDecodeError:
+                                continue
+
+        except Exception as e:
+            logger.error(f"QwenVL流式调用失败: {e}")
+            yield f"[ERROR: {str(e)}]"
+
+    def get_available_models(self) -> List[str]:
+        """获取可用模型列表"""
+        return self.SUPPORTED_MODELS
+
+    def close(self):
+        """关闭客户端"""
+        if self._client:
+            self._client.close()
+
+
+def create_qwen_client(
+    api_key: str,
+    model: str = "qwen3.5-flash",
+    base_url: str = "http://6.86.80.4:30080/v1",
+    **kwargs
+) -> QwenClient:
+    """便捷函数：创建Qwen客户端"""
+    config = LLMConfig(
+        provider=LLMProvider.QWEN,
+        model=model,
+        api_key=api_key,
+        base_url=base_url,
+        **kwargs
+    )
+    return QwenClient(config)
+
+
+def create_qwen_vl_client(
+    api_key: str,
+    model: str = "qwen3-vl-plus",
+    base_url: str = "http://6.86.80.4:30080/v1",
+    **kwargs
+) -> QwenVLClient:
+    """便捷函数：创建QwenVL客户端"""
+    config = LLMConfig(
+        provider=LLMProvider.QWEN_VL,
+        model=model,
+        api_key=api_key,
+        base_url=base_url,
+        **kwargs
+    )
+    return QwenVLClient(config)

backend/app/services/mock_data.py

@@ -0,0 +1,425 @@
+"""
+Mock数据服务 - 提供预设假数据供前后端对接测试
+"""
+
+from datetime import datetime
+from typing import Dict, List, Any
+import uuid
+
+# 预设法规文档列表
+MOCK_DOCUMENTS: List[Dict[str, Any]] = [
+    {
+        "id": "doc-001",
+        "name": "道路交通安全法.pdf",
+        "chunks": 156,
+        "status": "indexed",
+        "created_at": datetime(2026, 5, 10, 10, 0, 0),
+    },
+    {
+        "id": "doc-002",
+        "name": "机动车登记规定.docx",
+        "chunks": 89,
+        "status": "indexed",
+        "created_at": datetime(2026, 5, 10, 11, 0, 0),
+    },
+    {
+        "id": "doc-003",
+        "name": "电动自行车规范.pdf",
+        "chunks": 42,
+        "status": "indexed",
+        "created_at": datetime(2026, 5, 10, 12, 0, 0),
+    },
+    {
+        "id": "doc-004",
+        "name": "GB 38031-2020 电动汽车安全要求.pdf",
+        "chunks": 128,
+        "status": "indexed",
+        "created_at": datetime(2026, 5, 10, 13, 0, 0),
+    },
+    {
+        "id": "doc-005",
+        "name": "C-NCAP管理规则(2021版).pdf",
+        "chunks": 95,
+        "status": "indexed",
+        "created_at": datetime(2026, 5, 10, 14, 0, 0),
+    },
+]
+
+# 预设快捷问题
+MOCK_QUICK_QUESTIONS: List[Dict[str, str]] = [
+    {"id": "q1", "question": "电动自行车需要上牌照吗？", "category": "车辆登记"},
+    {"id": "q2", "question": "新能源汽车有哪些补贴政策？", "category": "新能源"},
+    {"id": "q3", "question": "车辆年检的规定是什么？", "category": "年检"},
+    {"id": "q4", "question": "驾驶证过期了怎么处理？", "category": "驾驶证"},
+]
+
+# 预设检索结果
+MOCK_RETRIEVAL_RESULTS: List[Dict[str, Any]] = [
+    {
+        "id": "chunk-001",
+        "score": 0.95,
+        "preview": "根据《道路交通安全法》第十八条规定，电动自行车经公安机关交通管理部门登记后，方可上道路行驶...",
+        "doc_name": "道路交通安全法",
+        "clause": "第十八条",
+        "content": "根据《道路交通安全法》第十八条规定，电动自行车经公安机关交通管理部门登记后，方可上道路行驶。电动自行车应当符合国家标准，最高设计车速不超过二十五公里每小时，整车质量不超过五十五千克。",
+    },
+    {
+        "id": "chunk-002",
+        "score": 0.88,
+        "preview": "电动自行车需符合GB17761-2018国家标准，包括最高车速、整车质量、脚踏骑行能力等要求...",
+        "doc_name": "电动自行车规范",
+        "clause": "第4条",
+        "content": "电动自行车需符合GB17761-2018国家标准。主要技术要求包括：最高设计车速不超过25km/h，整车质量不超过55kg，具有脚踏骑行能力，蓄电池标称电压不超过48V，电动机额定连续输出功率不超过400W。",
+    },
+    {
+        "id": "chunk-003",
+        "score": 0.82,
+        "preview": "机动车登记规定：初次申领机动车号牌、行驶证的，机动车所有人应当向住所地的车辆管理所申请注册登记...",
+        "doc_name": "机动车登记规定",
+        "clause": "第5条",
+        "content": "机动车登记规定：初次申领机动车号牌、行驶证的，机动车所有人应当向住所地的车辆管理所申请注册登记。申请注册登记的，应当提交机动车所有人的身份证明、购车发票等机动车来历证明、机动车整车出厂合格证明或者进口机动车进口凭证。",
+    },
+    {
+        "id": "chunk-004",
+        "score": 0.75,
+        "preview": "驾驶电动自行车上道路行驶，应当佩戴安全头盔，遵守道路交通安全法律法规...",
+        "doc_name": "道路交通安全法",
+        "clause": "第76条",
+        "content": "驾驶电动自行车上道路行驶，应当佩戴安全头盔，遵守道路交通安全法律法规。电动自行车不得逆向行驶，不得在机动车道内行驶，最高车速不得超过规定的限速。",
+    },
+    {
+        "id": "chunk-005",
+        "score": 0.68,
+        "preview": "电动汽车动力电池安全要求：电池系统发生热失控后，应在5分钟内不起火不爆炸...",
+        "doc_name": "GB 38031-2020",
+        "clause": "第7条",
+        "content": "电动汽车动力电池安全要求（GB 38031-2020）：电池系统发生热失控后，应在5分钟内不起火不爆炸，为乘员预留逃生时间。电池包需通过针刺、过充、短路等安全测试。",
+    },
+]
+
+# 预设RAG问答答案模板（按关键词匹配）
+MOCK_RAG_ANSWERS: Dict[str, Dict[str, Any]] = {
+    "电动自行车": {
+        "text": "根据《道路交通安全法》及相关规范，电动自行车上路需满足以下条件：\n\n1. 符合国家标准 GB17761-2018\n2. 经公安机关交通管理部门登记\n3. 最高设计车速不超过 25km/h\n4. 整车质量不超过 55kg\n5. 具有脚踏骑行能力\n6. 蓄电池标称电压不超过 48V\n\n行驶时还需佩戴安全头盔，不得逆向行驶或在机动车道内行驶。",
+        "retrieval_ids": ["chunk-001", "chunk-002", "chunk-004"],
+    },
+    "驾驶证": {
+        "text": "驾驶证申请流程如下：\n\n1. 到驾校报名并参加培训\n2. 通过科目一（理论考试）\n3. 通过科目二（场地驾驶技能考试）\n4. 通过科目三（道路驾驶技能考试）\n5. 通过科目四（安全文明驾驶常识考试）\n6. 领取驾驶证\n\n初次申领需到住所地车辆管理所申请注册登记。",
+        "retrieval_ids": ["chunk-003"],
+    },
+    "超速": {
+        "text": "超速处罚标准（根据《道路交通安全法》）：\n\n- 超速10%以下：警告\n- 超速10%-20%：罚款50-200元\n- 超速20%-50%：罚款200-500元，记3-6分\n- 超速50%以上：罚款500-2000元，记12分，可吊销驾驶证\n\n机动车驾驶人违反道路交通安全法律、法规将处警告或二十元以上二百元以下罚款。",
+        "retrieval_ids": ["chunk-001"],
+    },
+    "年检": {
+        "text": "车辆年检规定：\n\n- 小型私家车：6年内免检（每2年申领标志），6-10年每2年检验，10年以上每年检验\n- 车辆需携带行驶证、交强险保单\n- 检验项目：灯光、制动、排放等\n\n机动车所有人的住所迁出车辆管理所管辖区域的，需在登记证书上签注变更事项。",
+        "retrieval_ids": ["chunk-003"],
+    },
+    "电池": {
+        "text": "电动汽车电池安全标准（GB 38031-2020）：\n\n1. 热失控要求：电池系统发生热失控后，应在5分钟内不起火不爆炸，为乘员预留逃生时间\n2. 电池包需通过针刺、过充、短路等安全测试\n3. 充电系统应具备过充保护功能，当电池SOC达到100%时应自动停止充电\n4. 充电接口应符合GB/T 18487.1标准要求\n\n以上要求确保电动汽车的整车安全性。",
+        "retrieval_ids": ["chunk-005"],
+    },
+    "碰撞": {
+        "text": "正面碰撞测试要求（C-NCAP管理规则）：\n\n1. 正面100%重叠刚性壁障碰撞试验\n2. 碰撞速度：50km/h\n3. 试验后要求：\n   - 车门应能打开\n   - 燃油系统无泄漏\n   - 座椅及安全带功能正常\n\n此测试用于评估车辆在正面碰撞事故中对乘员的保护能力。",
+        "retrieval_ids": [],
+    },
+    "AEB": {
+        "text": "AEB（自动紧急制动系统）测试标准：\n\n1. 系统应在检测到前方障碍物时主动减速或停车\n2. 测试场景分为三种：\n   - 目标车静止\n   - 目标车移动\n   - 目标车制动\n3. AEB功能是C-NCAP评分的重要加分项\n\n该系统对提升车辆主动安全性能具有重要意义。",
+        "retrieval_ids": [],
+    },
+    "高速公路": {
+        "text": "高速公路安全距离规定：\n\n1. 车速超过100km/h时，与同车道前车保持100米以上距离\n2. 车速低于100km/h时，距离可适当缩短\n3. 执行紧急任务的警车、消防车、救护车、工程救险车不受行驶速度限制\n\n保持安全距离是预防追尾事故的关键措施。",
+        "retrieval_ids": [],
+    },
+}
+
+# 预设合规分析结果
+MOCK_COMPLIANCE_RESULT: Dict[str, Any] = {
+    "task_id": "task-001",
+    "dashboard": {
+        "score": 78,
+        "high_risk_count": 2,
+        "medium_risk_count": 1,
+        "low_risk_count": 0,
+        "need_fix_segments": 3,
+        "status": "warning",
+        "status_label": "需优化",
+    },
+    "segments": [
+        {
+            "id": 1,
+            "index": 1,
+            "intent": "车身结构设计",
+            "start_pos": 45,
+            "end_pos": 230,
+            "content": "车身采用高强度钢铝混合结构，A柱和B柱使用热成型钢板，厚度2.5mm。车顶结构设计满足GB 26112-2010抗压强度要求，正面碰撞能量吸收区域采用渐进式变形设计，确保碰撞时能量有效分散。",
+            "risk_level": "high",
+            "regulations": [
+                {
+                    "id": 1,
+                    "name": "GB 26112-2010",
+                    "clause": "第4.2条",
+                    "score": 0.95,
+                    "match_keyword": "车顶抗压强度",
+                    "category": "high",
+                    "full_content": "车顶结构应能承受相当于车辆整备质量1.5倍的载荷，载荷分布应均匀，试验后车顶变形量不超过规定值。",
+                },
+                {
+                    "id": 2,
+                    "name": "C-NCAP管理规则",
+                    "clause": "第3.1条",
+                    "score": 0.88,
+                    "match_keyword": "正面碰撞",
+                    "category": "high",
+                    "full_content": "正面碰撞试验速度为50km/h，碰撞后车门应能打开，燃油系统无泄漏，座椅及安全带功能正常。",
+                },
+                {
+                    "id": 3,
+                    "name": "GB 11551-2014",
+                    "clause": "第5条",
+                    "score": 0.72,
+                    "match_keyword": "碰撞能量吸收",
+                    "category": "medium",
+                    "full_content": "车辆正面碰撞时应有效保护乘员，碰撞能量应通过车身结构合理分散。",
+                },
+                {
+                    "id": 4,
+                    "name": "机动车安全技术条件",
+                    "clause": "第12条",
+                    "score": 0.58,
+                    "match_keyword": "A柱强度",
+                    "category": "medium",
+                    "full_content": "A柱应具备足够的抗变形能力，材料强度应符合相关标准要求。",
+                },
+            ],
+        },
+        {
+            "id": 2,
+            "index": 2,
+            "intent": "动力系统配置",
+            "start_pos": 298,
+            "end_pos": 425,
+            "content": "搭载永磁同步电机，最大功率150kW，峰值扭矩310Nm。电池组采用三元锂离子电池，容量75kWh，能量密度180Wh/kg。充电接口支持快充（30分钟充至80%）和慢充（8小时充满），符合GB/T 18487.1-2015标准。",
+            "risk_level": "medium",
+            "regulations": [
+                {
+                    "id": 5,
+                    "name": "GB/T 18487.1-2015",
+                    "clause": "第6条",
+                    "score": 0.94,
+                    "match_keyword": "充电接口标准",
+                    "category": "high",
+                    "full_content": "电动汽车传导充电接口应符合GB/T 18487.1标准要求，充电系统应具备过充保护功能。",
+                },
+                {
+                    "id": 6,
+                    "name": "GB/T 31484-2015",
+                    "clause": "第4条",
+                    "score": 0.85,
+                    "match_keyword": "电池能量密度",
+                    "category": "high",
+                    "full_content": "动力电池能量密度不低于120Wh/kg，电池系统需通过热失控测试。",
+                },
+                {
+                    "id": 7,
+                    "name": "新能源汽车生产企业准入",
+                    "clause": "第8条",
+                    "score": 0.65,
+                    "match_keyword": "电机功率",
+                    "category": "medium",
+                    "full_content": "驱动电机应符合相关技术标准，功率参数应在规定范围内。",
+                },
+                {
+                    "id": 8,
+                    "name": "电动汽车安全要求",
+                    "clause": "第7条",
+                    "score": 0.45,
+                    "match_keyword": "充电时间",
+                    "category": "low",
+                    "full_content": "充电系统应具备过充保护功能，当电池SOC达到100%时应自动停止充电。",
+                },
+            ],
+        },
+        {
+            "id": 3,
+            "index": 3,
+            "intent": "安全配置设计",
+            "start_pos": 570,
+            "end_pos": 725,
+            "content": "配备6个安全气囊（前排双气囊、侧气囊、侧气帘），采用预紧式安全带。ABS系统采用博世第9代ESP，具备碰撞预警功能（FCW）和自动紧急制动（AEB）。方向盘集成驾驶员疲劳监测摄像头。",
+            "risk_level": "low",
+            "regulations": [
+                {
+                    "id": 9,
+                    "name": "GB 27887-2011",
+                    "clause": "第5条",
+                    "score": 0.92,
+                    "match_keyword": "安全气囊",
+                    "category": "high",
+                    "full_content": "乘用车应配备驾驶员和乘客安全气囊，气囊系统应符合相关技术标准。",
+                },
+                {
+                    "id": 10,
+                    "name": "GB/T 26991-2011",
+                    "clause": "第3条",
+                    "score": 0.78,
+                    "match_keyword": "ABS系统",
+                    "category": "medium",
+                    "full_content": "车辆应配备防抱死制动系统，系统性能应符合相关标准要求。",
+                },
+                {
+                    "id": 11,
+                    "name": "C-NCAP管理规则",
+                    "clause": "第4.2条",
+                    "score": 0.71,
+                    "match_keyword": "AEB自动制动",
+                    "category": "medium",
+                    "full_content": "主动安全配置评分包含AEB功能，AEB系统应能有效检测障碍物并主动减速。",
+                },
+                {
+                    "id": 12,
+                    "name": "机动车运行安全技术条件",
+                    "clause": "第15条",
+                    "score": 0.38,
+                    "match_keyword": "疲劳监测",
+                    "category": "low",
+                    "full_content": "建议配备驾驶员状态监测系统，及时发现驾驶员疲劳或分心状态。",
+                },
+            ],
+        },
+    ],
+    "priority_actions": [
+        {
+            "regulation": "GB 26112-2010 第4.2条",
+            "issue": "缺少车顶抗压强度测试数据",
+            "suggestion": "补充车顶抗压强度具体测试数据，确保满足1.5倍整备质量载荷要求",
+            "severity": "high",
+        },
+        {
+            "regulation": "GB/T 31484-2015 第4条",
+            "issue": "缺少电池热失控测试报告",
+            "suggestion": "补充电池热失控测试报告，验证5分钟内不起火不爆炸",
+            "severity": "high",
+        },
+        {
+            "regulation": "C-NCAP管理规则 第3.1条",
+            "issue": "缺少碰撞后车门开启性能数据",
+            "suggestion": "提供碰撞后车门开启性能测试数据",
+            "severity": "medium",
+        },
+    ],
+}
+
+# 预设合规对话响应模板
+MOCK_COMPLIANCE_CHAT_RESPONSES: Dict[str, Dict[str, str]] = {
+    "车身结构设计": {
+        "compliance": "根据当前分析，车身结构设计部分存在以下合规问题：\n\n1. GB 26112-2010要求车顶承受1.5倍整备质量载荷，目前设计声明满足要求但缺少测试数据\n2. C-NCAP正面碰撞后车门应能打开，需提供碰撞测试报告\n\n建议补充相关测试数据以提升合规评分。",
+        "interpretation": "GB 26112-2010 第4.2条具体要求解读：\n\n车顶抗压强度测试是车辆被动安全的重要指标。该标准要求车顶结构能够承受相当于车辆整备质量1.5倍的均匀分布载荷，试验后车顶变形量不得超过规定限值。\n\n热成型钢板（22MnB5材料）抗拉强度约1500-1650 MPa，理论上能满足要求，但需通过实际测试验证。",
+        "suggestion": "针对车身结构设计的修改建议：\n\n1. 补充车顶抗压强度测试报告\n2. 提供A柱材料认证证书\n3. 完善正面碰撞能量吸收设计说明\n4. 添加碰撞后车门开启性能数据\n\n这些补充材料可有效提升合规评分。",
+    },
+    "动力系统配置": {
+        "compliance": "动力系统配置整体合规性良好，主要检查点：\n\n1. 电池能量密度180Wh/kg超过最低要求120Wh/kg ✓\n2. 充电接口符合GB/T 18487.1标准 ✓\n3. 快充30分钟充至80%符合行业标准 ✓\n\n需补充电池热失控测试报告。",
+        "interpretation": "GB/T 31484-2015对动力电池的要求解读：\n\n1. 能量密度：不低于120Wh/kg（您的设计180Wh/kg满足要求）\n2. 循环寿命：不少于1000次循环后容量保持率≥80%\n3. 安全测试：需通过针刺、过充、短路等测试\n\n建议补充循环寿命测试数据。",
+        "suggestion": "动力系统配置改进建议：\n\n1. 补充电池热失控测试报告\n2. 提供循环寿命测试数据\n3. 添加充电系统过充保护功能说明\n4. 完善电池管理系统（BMS）技术文档",
+    },
+    "安全配置设计": {
+        "compliance": "安全配置设计合规性评估：\n\n1. 安全气囊配置满足GB 27887-2011要求 ✓\n2. ABS/ESP系统符合标准 ✓\n3. AEB功能是C-NCAP加分项 ✓\n\n驾驶员疲劳监测是建议配置，不强制要求。",
+        "interpretation": "C-NCAP主动安全评分规则解读：\n\nAEB（自动紧急制动）系统是C-NCAP评分的重要加分项，最高可获得额外加分。测试场景包括：\n- 目标车静止场景\n- 目标车移动场景\n- 目标车制动场景\n\n建议完善AEB系统测试数据以获取更高评分。",
+        "suggestion": "安全配置优化建议：\n\n1. 提供AEB系统测试数据\n2. 补充FCW预警功能测试报告\n3. 添加安全气囊展开时间数据\n4. 完善驾驶员疲劳监测系统说明（如有）",
+    },
+}
+
+# 预设系统统计数据
+MOCK_SYSTEM_STATS: Dict[str, int] = {
+    "docs": 5,
+    "chunks": 510,
+    "vectors": 510,
+    "segments": 0,
+}
+
+# 预设系统配置
+MOCK_SYSTEM_CONFIG: Dict[str, Any] = {
+    "llm": {
+        "model": "qwen-max",
+    },
+    "embedding": {
+        "model": "text-embedding-v3",
+        "dimension": 1536,
+    },
+    "milvus": {
+        "host": "localhost",
+        "port": 19530,
+    },
+    "retrieval": {
+        "vector_top_k": 10,
+        "final_top_k": 5,
+    },
+}
+
+
+def get_mock_documents() -> List[Dict[str, Any]]:
+    """获取预设法规文档列表"""
+    return MOCK_DOCUMENTS
+
+
+def get_mock_quick_questions() -> List[Dict[str, str]]:
+    """获取预设快捷问题"""
+    return MOCK_QUICK_QUESTIONS
+
+
+def get_mock_retrieval(query: str, top_k: int = 5) -> List[Dict[str, Any]]:
+    """根据查询关键词返回预设检索结果"""
+    results = []
+    for keyword, data in MOCK_RAG_ANSWERS.items():
+        if keyword in query:
+            for retrieval_id in data.get("retrieval_ids", []):
+                for item in MOCK_RETRIEVAL_RESULTS:
+                    if item["id"] == retrieval_id:
+                        results.append({
+                            "id": item["id"],
+                            "score": item["score"],
+                            "preview": item["preview"],
+                            "doc_name": item["doc_name"],
+                            "clause": item["clause"],
+                        })
+                        break
+    if not results:
+        results = MOCK_RETRIEVAL_RESULTS[:top_k]
+    return results[:top_k]
+
+
+def get_mock_rag_answer(query: str) -> str:
+    """根据查询关键词返回预设答案"""
+    for keyword, data in MOCK_RAG_ANSWERS.items():
+        if keyword in query:
+            return data["text"]
+    return "抱歉，暂未找到与您问题直接相关的法规内容。请尝试更具体的问题，或联系交通管理部门获取详细信息。\n\n您可以尝试询问：电动自行车、驾驶证、超速处罚、年检、电池安全、碰撞测试、AEB系统、高速公路规则等话题。"
+
+
+def get_mock_compliance_result(task_id: str) -> Dict[str, Any]:
+    """获取预设合规分析结果"""
+    result = MOCK_COMPLIANCE_RESULT.copy()
+    result["task_id"] = task_id
+    return result
+
+
+def get_mock_compliance_chat_response(intent: str, query: str) -> str:
+    """获取预设合规对话响应"""
+    responses = MOCK_COMPLIANCE_CHAT_RESPONSES.get(intent, {})
+    if "合规" in query or "符合" in query:
+        return responses.get("compliance", "根据相关法规分析，该段落的合规性需进一步评估。")
+    elif "解读" in query or "什么" in query or "如何" in query:
+        return responses.get("interpretation", "法规要求详细解读如下...")
+    elif "修改" in query or "建议" in query or "完善" in query:
+        return responses.get("suggestion", "建议进行以下修改以提升合规性...")
+    return f"关于您的问题，{intent}部分涉及多条相关法规。您可以进一步询问合规性评估或修改建议。"
+
+
+def generate_task_id() -> str:
+    """生成任务ID"""
+    return f"task-{uuid.uuid4().hex[:8]}"
+
+
+def generate_doc_id() -> str:
+    """生成文档ID"""
+    return f"doc-{uuid.uuid4().hex[:8]}"

backend/app/services/parser/__init__.py

@@ -0,0 +1,7 @@
+# src/services/parser/__init__.py
+"""文档解析服务"""
+
+from .pdf_parser import PDFParser
+from .docx_parser import DocxParser
+
+__all__ = ["PDFParser", "DocxParser"]

backend/app/services/parser/docx_parser.py

@@ -0,0 +1,287 @@
+# src/services/parser/docx_parser.py
+"""Word文档解析 - 使用python-docx"""
+
+from docx import Document
+from docx.enum.text import WD_ALIGN_PARAGRAPH
+from typing import List, Dict, Optional
+from dataclasses import dataclass, field
+from loguru import logger
+import re
+
+
+@dataclass
+class DocxParagraph:
+    """段落内容"""
+    text: str
+    level: int = 0  # 标题级别，0表示正文
+    is_list: bool = False
+    list_number: Optional[str] = None
+
+
+@dataclass
+class DocxTable:
+    """表格内容"""
+    rows: List[List[str]]
+    markdown: str = ""
+
+
+@dataclass
+class DocxDocumentContent:
+    """Word文档完整内容"""
+    file_path: str
+    paragraphs: List[DocxParagraph]
+    tables: List[DocxTable]
+    metadata: Dict[str, str] = field(default_factory=dict)
+    markdown_text: str = ""
+
+
+class DocxParser:
+    """Word文档解析器 - 基于python-docx"""
+
+    def __init__(self):
+        self.document = None
+
+    def parse(self, file_path: str) -> DocxDocumentContent:
+        """
+        解析Word文档
+
+        Args:
+            file_path: Word文档路径
+
+        Returns:
+            DocxDocumentContent: 解析后的文档内容
+        """
+        logger.info(f"开始解析Word文档: {file_path}")
+
+        try:
+            self.document = Document(file_path)
+            doc_content = DocxDocumentContent(
+                file_path=file_path,
+                paragraphs=[],
+                tables=[]
+            )
+
+            # 提取文档元数据
+            doc_content.metadata = self._extract_metadata()
+
+            # 提取段落
+            doc_content.paragraphs = self._extract_paragraphs()
+
+            # 提取表格
+            doc_content.tables = self._extract_tables()
+
+            # 生成Markdown格式文本
+            doc_content.markdown_text = self._generate_markdown(doc_content)
+
+            logger.success(f"Word文档解析完成，共{len(doc_content.paragraphs)}个段落")
+
+            return doc_content
+
+        except Exception as e:
+            logger.error(f"Word文档解析失败: {e}")
+            raise
+
+    def _extract_metadata(self) -> Dict[str, str]:
+        """提取文档元数据"""
+        metadata = {}
+        try:
+            core_props = self.document.core_properties
+            metadata = {
+                "title": core_props.title or "",
+                "author": core_props.author or "",
+                "subject": core_props.subject or "",
+                "keywords": core_props.keywords or "",
+                "created": str(core_props.created) if core_props.created else "",
+                "modified": str(core_props.modified) if core_props.modified else "",
+            }
+        except Exception as e:
+            logger.warning(f"提取元数据失败: {e}")
+        return metadata
+
+    def _extract_paragraphs(self) -> List[DocxParagraph]:
+        """提取所有段落"""
+        paragraphs = []
+
+        for para in self.document.paragraphs:
+            text = para.text.strip()
+            if not text:
+                continue
+
+            # 判断标题级别
+            level = self._get_paragraph_level(para)
+
+            # 判断是否是列表项
+            is_list, list_number = self._detect_list_item(para)
+
+            paragraph = DocxParagraph(
+                text=text,
+                level=level,
+                is_list=is_list,
+                list_number=list_number
+            )
+            paragraphs.append(paragraph)
+
+        return paragraphs
+
+    def _get_paragraph_level(self, para) -> int:
+        """
+        判断段落标题级别
+
+        Returns:
+            int: 标题级别，0表示正文
+        """
+        # 方法1：检查段落样式
+        style_name = para.style.name if para.style else ""
+
+        if "Heading" in style_name or "标题" in style_name:
+            # 从样式名称中提取级别
+            match = re.search(r'Heading\s*(\d)|标题\s*(\d)', style_name)
+            if match:
+                level = int(match.group(1) or match.group(2))
+                return level
+
+        # 方法2：检查段落格式（字号）
+        # 标题通常字号较大
+        if para.paragraph_format:
+            # 可以根据字号判断，这里简化处理
+            pass
+
+        # 方法3：根据内容模式判断（法规文档特征）
+        text = para.text.strip()
+
+        # 第一章、第X章 -> 二级标题
+        if re.match(r'^第[一二三四五六七八九十百]+章\s', text):
+            return 2
+        # 第X节 -> 三级标题
+        elif re.match(r'^第[一二三四五六七八九十百]+节\s', text):
+            return 3
+        # 第X条 -> 四级标题
+        elif re.match(r'^第[一二三四五六七八九十百]+条\s', text):
+            return 4
+
+        return 0  # 正文
+
+    def _detect_list_item(self, para) -> tuple[bool, Optional[str]]:
+        """检测是否是列表项"""
+        text = para.text.strip()
+
+        # 数字列表：1.、2.、（1）、[1]等
+        if re.match(r'^[\d]+[.、)\]]\s', text):
+            match = re.match(r'^([\d]+[.、)\]])\s', text)
+            return True, match.group(1) if match else None
+
+        # 中文数字列表：一、二、（一）等
+        if re.match(r'^[一二三四五六七八九十]+[、.)]\s', text):
+            match = re.match(r'^([一二三四五六七八九十]+[、.)])\s', text)
+            return True, match.group(1) if match else None
+
+        # 检查段落格式中的列表编号
+        if para.paragraph_format and hasattr(para.paragraph_format, 'left_indent'):
+            # 有缩进的可能是列表项
+            pass
+
+        return False, None
+
+    def _extract_tables(self) -> List[DocxTable]:
+        """提取所有表格"""
+        tables = []
+
+        for table in self.document.tables:
+            rows = []
+            for row in table.rows:
+                cells = []
+                for cell in row.cells:
+                    cells.append(cell.text.strip())
+                rows.append(cells)
+
+            # 转换为Markdown表格
+            markdown = self._table_to_markdown(rows)
+
+            table_content = DocxTable(rows=rows, markdown=markdown)
+            tables.append(table_content)
+
+        return tables
+
+    def _table_to_markdown(self, rows: List[List[str]]) -> str:
+        """将表格转换为Markdown格式"""
+        if not rows or len(rows) < 1:
+            return ""
+
+        lines = []
+
+        # 表头
+        if len(rows) >= 1:
+            header = rows[0]
+            lines.append("| " + " | ".join(cell for cell in header) + " |")
+            lines.append("| " + " | ".join("---" for _ in header) + " |")
+
+        # 数据行
+        for row in rows[1:]:
+            lines.append("| " + " | ".join(cell for cell in row) + " |")
+
+        return "\n".join(lines)
+
+    def _generate_markdown(self, doc_content: DocxDocumentContent) -> str:
+        """生成Markdown格式文本"""
+        lines = []
+
+        # 文档标题
+        title = doc_content.metadata.get("title", "")
+        if title:
+            lines.append(f"# {title}\n")
+        else:
+            # 从第一个段落获取标题（如果是标题样式）
+            for para in doc_content.paragraphs[:5]:
+                if para.level == 1:
+                    lines.append(f"# {para.text}\n")
+                    break
+            else:
+                lines.append(f"# {doc_content.file_path}\n")
+
+        # 元数据信息
+        lines.append("\n## 文档信息\n")
+        for key, value in doc_content.metadata.items():
+            if value:
+                lines.append(f"- **{key}**: {value}")
+
+        # 正文内容
+        lines.append("\n## 正文\n")
+
+        table_index = 0
+        for para in doc_content.paragraphs:
+            if para.level > 0:
+                # 标题
+                prefix = "#" * para.level
+                lines.append(f"\n{prefix} {para.text}\n")
+            elif para.is_list:
+                # 列表项
+                lines.append(f"- {para.text}")
+            else:
+                # 正文
+                lines.append(para.text)
+
+        # 添加表格
+        if doc_content.tables:
+            lines.append("\n## 表格\n")
+            for i, table in enumerate(doc_content.tables):
+                lines.append(f"\n### 表格 {i + 1}\n")
+                lines.append(table.markdown + "\n")
+
+        return "\n".join(lines)
+
+    def parse_to_markdown(self, file_path: str) -> str:
+        """直接解析并返回Markdown文本"""
+        doc_content = self.parse(file_path)
+        return doc_content.markdown_text
+
+
+def parse_docx(file_path: str) -> DocxDocumentContent:
+    """便捷函数：解析Word文档"""
+    parser = DocxParser()
+    return parser.parse(file_path)
+
+
+def parse_docx_to_markdown(file_path: str) -> str:
+    """便捷函数：解析Word并返回Markdown"""
+    parser = DocxParser()
+    return parser.parse_to_markdown(file_path)

backend/app/services/parser/mineru_parser.py

@@ -0,0 +1,204 @@
+# src/services/parser/mineru_parser.py
+"""MinerU多模态PDF解析 - 版面感知解析"""
+
+from typing import Optional, Dict
+from dataclasses import dataclass, field
+from loguru import logger
+import os
+
+
+@dataclass
+class MinerUResult:
+    """MinerU解析结果"""
+    file_path: str
+    markdown_text: str
+    metadata: Dict[str, str] = field(default_factory=dict)
+    success: bool = True
+    error_message: str = ""
+
+
+class MinerUParser:
+    """
+    MinerU多模态PDF解析器
+
+    MinerU (magic-pdf) 是一个开源的高质量PDF解析工具，
+    支持版面感知解析，能够识别文档中的标题、正文、表格、图片等元素，
+    并输出结构化的Markdown格式。
+
+    GitHub: https://github.com/opendatalab/MinerU
+    """
+
+    def __init__(self):
+        self.available = self._check_mineru_available()
+
+    def _check_mineru_available(self) -> bool:
+        """检查MinerU是否可用"""
+        try:
+            from magic_pdf.pipe.UNIPipe import UNIPipe
+            return True
+        except ImportError:
+            logger.warning("MinerU (magic-pdf) 未安装，请运行: pip install magic-pdf[full]")
+            return False
+
+    def parse(self, file_path: str, output_dir: Optional[str] = None) -> MinerUResult:
+        """
+        使用MinerU解析PDF文档
+
+        Args:
+            file_path: PDF文件路径
+            output_dir: 输出目录（可选，用于保存解析产物）
+
+        Returns:
+            MinerUResult: 解析结果
+        """
+        logger.info(f"尝试使用MinerU解析: {file_path}")
+
+        if not self.available:
+            return MinerUResult(
+                file_path=file_path,
+                markdown_text="",
+                success=False,
+                error_message="MinerU未安装"
+            )
+
+        try:
+            from magic_pdf.pipe.UNIPipe import UNIPipe
+            from magic_pdf.libs.MakeContentConfig import DropMode
+
+            # 设置输出目录
+            if output_dir is None:
+                output_dir = os.path.dirname(file_path)
+
+            # 创建解析管道
+            # OCR模式可以根据PDF类型选择
+            # auto: 自动判断是否需要OCR
+            # txt: 纯文本PDF（无OCR）
+            # ocr: 扫描件PDF（OCR）
+            pipe = UNIPipe(file_path, output_dir)
+
+            # 执行解析
+            # pipe_mk() 返回Markdown格式文本
+            markdown_content = pipe.pipe_mk()
+
+            logger.success(f"MinerU解析成功")
+
+            return MinerUResult(
+                file_path=file_path,
+                markdown_text=markdown_content,
+                metadata=self._extract_metadata(pipe),
+                success=True
+            )
+
+        except Exception as e:
+            logger.error(f"MinerU解析失败: {e}")
+            return MinerUResult(
+                file_path=file_path,
+                markdown_text="",
+                success=False,
+                error_message=str(e)
+            )
+
+    def _extract_metadata(self, pipe) -> Dict[str, str]:
+        """从解析管道提取元数据"""
+        metadata = {}
+        try:
+            # MinerU解析管道中可能包含的元数据信息
+            if hasattr(pipe, 'pdf_mid_data') and pipe.pdf_mid_data:
+                mid_data = pipe.pdf_mid_data
+                # 提取可能的元数据字段
+                metadata = {
+                    "page_count": str(mid_data.get("page_count", "")),
+                    "language": str(mid_data.get("language", "")),
+                    "is_scanned": str(mid_data.get("is_scanned", "")),
+                }
+        except Exception as e:
+            logger.warning(f"提取MinerU元数据失败: {e}")
+
+        return metadata
+
+    def parse_to_markdown(self, file_path: str) -> str:
+        """直接解析并返回Markdown文本"""
+        result = self.parse(file_path)
+        return result.markdown_text if result.success else ""
+
+
+class ParserOrchestrator:
+    """
+    解析服务编排 - 按优先级选择解析器
+
+    解析策略：
+    1. 优先尝试MinerU（版面感知能力强）
+    2. MinerU失败时回退到基础PyMuPDF解析
+    """
+
+    def __init__(self):
+        from .pdf_parser import PDFParser
+        self.mineru_parser = MinerUParser()
+        self.pdf_parser = PDFParser()
+        self.mineru_available = self.mineru_parser.available
+
+    def parse_pdf(self, file_path: str, prefer_mineru: bool = True) -> str:
+        """
+        解析PDF文档，按优先级选择解析器
+
+        Args:
+            file_path: PDF文件路径
+            prefer_mineru: 是否优先使用MinerU
+
+        Returns:
+            str: Markdown格式文本
+        """
+        markdown_text = ""
+
+        if prefer_mineru and self.mineru_available:
+            # 优先尝试MinerU
+            result = self.mineru_parser.parse(file_path)
+            if result.success:
+                markdown_text = result.markdown_text
+                logger.info("使用MinerU解析成功")
+                return markdown_text
+            else:
+                logger.warning(f"MinerU解析失败，回退到PyMuPDF: {result.error_message}")
+
+        # 回退到PyMuPDF基础解析
+        logger.info("使用PyMuPDF基础解析")
+        markdown_text = self.pdf_parser.parse_to_markdown(file_path)
+
+        return markdown_text
+
+    def parse_docx(self, file_path: str) -> str:
+        """解析Word文档"""
+        from .docx_parser import DocxParser
+        docx_parser = DocxParser()
+        return docx_parser.parse_to_markdown(file_path)
+
+    def parse(self, file_path: str) -> str:
+        """
+        根据文件类型选择解析器
+
+        Args:
+            file_path: 文件路径
+
+        Returns:
+            str: Markdown格式文本
+        """
+        ext = os.path.splitext(file_path)[1].lower()
+
+        if ext == ".pdf":
+            return self.parse_pdf(file_path)
+        elif ext in [".docx", ".doc"]:
+            return self.parse_docx(file_path)
+        else:
+            raise ValueError(f"不支持的文件类型: {ext}")
+
+
+def parse_with_mineru(file_path: str) -> MinerUResult:
+    """便捷函数：使用MinerU解析"""
+    parser = MinerUParser()
+    return parser.parse(file_path)
+
+
+def parse_pdf_smart(file_path: str) -> str:
+    """便捷函数：智能解析PDF（自动选择最佳解析器）"""
+    orchestrator = ParserOrchestrator()
+    return orchestrator.parse_pdf(file_path)

backend/app/services/parser/pdf_parser.py

@@ -0,0 +1,268 @@
+# src/services/parser/pdf_parser.py
+"""PDF文档解析 - 使用PyMuPDF基础解析"""
+
+import fitz  # PyMuPDF
+from typing import List, Dict, Optional, Tuple
+from dataclasses import dataclass, field
+from loguru import logger
+import re
+
+
+@dataclass
+class PDFPageContent:
+    """PDF页面内容"""
+    page_number: int
+    text: str
+    tables: List[str] = field(default_factory=list)
+    images: List[str] = field(default_factory=list)  # 图片路径列表
+    blocks: List[Dict] = field(default_factory=list)
+
+
+@dataclass
+class PDFDocumentContent:
+    """PDF文档完整内容"""
+    file_path: str
+    total_pages: int
+    pages: List[PDFPageContent]
+    metadata: Dict[str, str] = field(default_factory=dict)
+    markdown_text: str = ""
+
+
+class PDFParser:
+    """PDF文档解析器 - 基于PyMuPDF"""
+
+    def __init__(self):
+        self.pdf = None
+
+    def parse(self, file_path: str, extract_tables: bool = True, extract_images: bool = False) -> PDFDocumentContent:
+        """
+        解析PDF文档
+
+        Args:
+            file_path: PDF文件路径
+            extract_tables: 是否提取表格
+            extract_images: 是否提取图片
+
+        Returns:
+            PDFDocumentContent: 解析后的文档内容
+        """
+        logger.info(f"开始解析PDF文档: {file_path}")
+
+        try:
+            self.pdf = fitz.open(file_path)
+            doc_content = PDFDocumentContent(
+                file_path=file_path,
+                total_pages=self.pdf.page_count,
+                pages=[]
+            )
+
+            # 提取文档元数据
+            doc_content.metadata = self._extract_metadata()
+
+            # 逐页解析
+            for page_num in range(self.pdf.page_count):
+                page = self.pdf[page_num]
+                page_content = self._parse_page(page, page_num + 1, extract_tables, extract_images)
+                doc_content.pages.append(page_content)
+
+            # 生成Markdown格式文本
+            doc_content.markdown_text = self._generate_markdown(doc_content)
+
+            self.pdf.close()
+            logger.success(f"PDF解析完成，共{doc_content.total_pages}页")
+
+            return doc_content
+
+        except Exception as e:
+            logger.error(f"PDF解析失败: {e}")
+            raise
+
+    def _extract_metadata(self) -> Dict[str, str]:
+        """提取PDF元数据"""
+        metadata = {}
+        try:
+            meta = self.pdf.metadata
+            metadata = {
+                "title": meta.get("title", ""),
+                "author": meta.get("author", ""),
+                "subject": meta.get("subject", ""),
+                "keywords": meta.get("keywords", ""),
+                "creator": meta.get("creator", ""),
+                "producer": meta.get("producer", ""),
+                "creation_date": meta.get("creationDate", ""),
+                "mod_date": meta.get("modDate", ""),
+            }
+        except Exception as e:
+            logger.warning(f"提取元数据失败: {e}")
+        return metadata
+
+    def _parse_page(self, page: fitz.Page, page_num: int,
+                   extract_tables: bool, extract_images: bool) -> PDFPageContent:
+        """解析单页内容"""
+        page_content = PDFPageContent(page_number=page_num, text="")
+
+        # 提取文本块（保留结构）
+        blocks = page.get_text("dict", flags=fitz.TEXT_PRESERVE_WHITESPACE)["blocks"]
+        page_content.blocks = blocks
+
+        # 提取纯文本
+        text = page.get_text("text", flags=fitz.TEXT_PRESERVE_WHITESPACE)
+        page_content.text = text.strip()
+
+        # 提取表格（使用PyMuPDF的表格提取功能）
+        if extract_tables:
+            tables = self._extract_tables_from_page(page)
+            page_content.tables = tables
+
+        # 提取图片
+        if extract_images:
+            images = self._extract_images_from_page(page, page_num)
+            page_content.images = images
+
+        return page_content
+
+    def _extract_tables_from_page(self, page: fitz.Page) -> List[str]:
+        """
+        从页面提取表格（基于文本块分析）
+        注意：PyMuPDF基础版表格提取能力有限，复杂表格建议使用MinerU
+        """
+        tables = []
+
+        try:
+            # 使用PyMuPDF的表格提取方法（2.4+版本）
+            # 对于更复杂的表格，需要在mineru_parser中使用更高级的方法
+            tabs = page.find_tables()
+            if tabs:
+                for tab in tabs:
+                    table_text = tab.extract()
+                    # 将表格转换为Markdown格式
+                    markdown_table = self._table_to_markdown(table_text)
+                    tables.append(markdown_table)
+
+        except AttributeError:
+            # 旧版本PyMuPDF没有表格提取功能
+            logger.warning("PyMuPDF版本不支持表格提取，请升级到2.4+版本")
+        except Exception as e:
+            logger.warning(f"表格提取失败: {e}")
+
+        return tables
+
+    def _table_to_markdown(self, table_data: List[List[str]]) -> str:
+        """将表格数据转换为Markdown格式"""
+        if not table_data or len(table_data) < 1:
+            return ""
+
+        lines = []
+        # 表头
+        if len(table_data) >= 1:
+            header = table_data[0]
+            lines.append("| " + " | ".join(str(cell).strip() for cell in header) + " |")
+            lines.append("| " + " | ".join("---" for _ in header) + " |")
+
+        # 数据行
+        for row in table_data[1:]:
+            lines.append("| " + " | ".join(str(cell).strip() for cell in row) + " |")
+
+        return "\n".join(lines)
+
+    def _extract_images_from_page(self, page: fitz.Page, page_num: int) -> List[str]:
+        """提取页面图片"""
+        images = []
+        # 图片提取功能（可选实现）
+        # 这里仅记录图片信息，实际图片需要额外保存
+        try:
+            image_list = page.get_images()
+            for img_index, img in enumerate(image_list):
+                xref = img[0]
+                images.append(f"image_p{page_num}_i{img_index}_xref{xref}")
+        except Exception as e:
+            logger.warning(f"图片提取失败: {e}")
+        return images
+
+    def _generate_markdown(self, doc_content: PDFDocumentContent) -> str:
+        """生成Markdown格式文本"""
+        lines = []
+
+        # 文档标题
+        title = doc_content.metadata.get("title", "")
+        if title:
+            lines.append(f"# {title}\n")
+        else:
+            lines.append(f"# {doc_content.file_path}\n")
+
+        # 元数据信息
+        lines.append("\n## 文档信息\n")
+        for key, value in doc_content.metadata.items():
+            if value and key in ["author", "subject", "keywords", "creation_date"]:
+                lines.append(f"- **{key}**: {value}")
+
+        # 正文内容
+        lines.append("\n## 正文\n")
+
+        for page in doc_content.pages:
+            # 页码标记
+            lines.append(f"\n---\n**第 {page.page_number} 页**\n")
+
+            # 处理文本内容，识别标题结构
+            text = self._process_page_text(page.text, page.blocks)
+            lines.append(text)
+
+            # 添加表格
+            for table in page.tables:
+                lines.append("\n" + table + "\n")
+
+        return "\n".join(lines)
+
+    def _process_page_text(self, text: str, blocks: List[Dict]) -> str:
+        """处理页面文本，识别标题结构"""
+        # 基于字体大小识别标题
+        processed_text = text
+
+        # 尝试识别标题（基于字号）
+        # 法规文档通常有明确的层级结构：章、节、条
+        processed_text = self._detect_headers(text, blocks)
+
+        return processed_text
+
+    def _detect_headers(self, text: str, blocks: List[Dict]) -> str:
+        """检测并标记标题（基于字号或内容模式）"""
+        lines = text.split("\n")
+        processed_lines = []
+
+        for line in lines:
+            line = line.strip()
+            if not line:
+                continue
+
+            # 法规标题模式检测
+            # 第一章、第X章、第X节、第X条等
+            if re.match(r'^第[一二三四五六七八九十百]+章\s', line):
+                processed_lines.append(f"\n## {line}\n")
+            elif re.match(r'^第[一二三四五六七八九十百]+节\s', line):
+                processed_lines.append(f"\n### {line}\n")
+            elif re.match(r'^第[一二三四五六七八九十百]+条\s', line):
+                processed_lines.append(f"\n#### {line}\n")
+            elif re.match(r'^[一二三四五六七八九十]+\s*[、.]', line):
+                # 条款子项
+                processed_lines.append(f"- {line}")
+            else:
+                processed_lines.append(line)
+
+        return "\n".join(processed_lines)
+
+    def parse_to_markdown(self, file_path: str) -> str:
+        """直接解析并返回Markdown文本"""
+        doc_content = self.parse(file_path)
+        return doc_content.markdown_text
+
+
+def parse_pdf(file_path: str, **kwargs) -> PDFDocumentContent:
+    """便捷函数：解析PDF文档"""
+    parser = PDFParser()
+    return parser.parse(file_path, **kwargs)
+
+
+def parse_pdf_to_markdown(file_path: str) -> str:
+    """便捷函数：解析PDF并返回Markdown"""
+    parser = PDFParser()
+    return parser.parse_to_markdown(file_path)

backend/app/services/rag/__init__.py

@@ -0,0 +1,12 @@
+# src/services/rag/__init__.py
+"""RAG服务模块"""
+
+from .retriever import Retriever, retrieve_regulations
+from .context_builder import ContextBuilder, build_rag_context
+from .prompt_templates import PromptTemplates, get_prompt_template
+
+__all__ = [
+    "Retriever", "retrieve_regulations",
+    "ContextBuilder", "build_rag_context",
+    "PromptTemplates", "get_prompt_template"
+]

backend/app/services/rag/context_builder.py

@@ -0,0 +1,230 @@
+# src/services/rag/context_builder.py
+"""RAG上下文构建服务 - 构建LLM输入上下文"""
+
+from typing import List, Dict, Optional
+from dataclasses import dataclass
+from loguru import logger
+
+from .retriever import RetrievedDocument
+from app.config.settings import settings
+
+
+@dataclass
+class RAGContext:
+    """RAG构建的上下文"""
+    system_prompt: str
+    context_text: str
+    user_query: str
+    total_tokens: int
+    sources: List[Dict]
+    truncated: bool = False
+
+
+class ContextBuilder:
+    """
+    RAG上下文构建器
+
+    功能：
+    - 格式化检索结果为上下文文本
+    - 控制上下文长度（token限制）
+    - 构建完整的LLM输入格式
+    """
+
+    def __init__(
+        self,
+        max_context_tokens: int = None,
+        include_metadata: bool = True,
+        citation_format: str = "【条款{clause}】"
+    ):
+        """
+        初始化上下文构建器
+
+        Args:
+            max_context_tokens: 最大上下文token数
+            include_metadata: 是否包含元数据（文档名、条款号等）
+            citation_format: 引用格式模板
+        """
+        self.max_context_tokens = max_context_tokens or settings.rag_max_context_tokens
+        self.include_metadata = include_metadata
+        self.citation_format = citation_format
+
+        logger.info(f"上下文构建器初始化: max_tokens={self.max_context_tokens}")
+
+    def build(
+        self,
+        query: str,
+        documents: List[RetrievedDocument],
+        system_prompt: Optional[str] = None,
+        max_tokens: Optional[int] = None
+    ) -> RAGContext:
+        """
+        构建RAG上下文
+
+        Args:
+            query: 用户查询
+            documents: 检索到的文档列表
+            system_prompt: 系统提示词（可选）
+            max_tokens: 最大token数（可选，覆盖默认值）
+
+        Returns:
+            RAGContext: 构建的上下文对象
+        """
+        max_tokens = max_tokens or self.max_context_tokens
+
+        # 格式化文档内容
+        context_text, sources, truncated = self._format_documents(
+            documents,
+            max_tokens
+        )
+
+        # 构建系统提示词
+        system_prompt = system_prompt or self._default_system_prompt()
+
+        # 估算总token数
+        total_tokens = self._estimate_tokens(system_prompt + context_text + query)
+
+        logger.info(f"上下文构建完成: {len(documents)}条文档, {total_tokens}tokens, truncated={truncated}")
+
+        return RAGContext(
+            system_prompt=system_prompt,
+            context_text=context_text,
+            user_query=query,
+            total_tokens=total_tokens,
+            sources=sources,
+            truncated=truncated
+        )
+
+    def _format_documents(
+        self,
+        documents: List[RetrievedDocument],
+        max_tokens: int
+    ) -> tuple:
+        """
+        格式化文档内容
+
+        Args:
+            documents: 文档列表
+            max_tokens: 最大token数
+
+        Returns:
+            (context_text, sources, truncated)
+        """
+        context_parts = []
+        sources = []
+        current_tokens = 0
+        truncated = False
+
+        for i, doc in enumerate(documents):
+            # 格式化单个文档
+            formatted = self._format_single_doc(doc, i + 1)
+
+            # 估算token数
+            doc_tokens = self._estimate_tokens(formatted)
+
+            # 检查是否超出限制
+            if current_tokens + doc_tokens > max_tokens:
+                truncated = True
+                logger.warning(f"上下文截断: 已达到{max_tokens}tokens限制")
+                break
+
+            context_parts.append(formatted)
+            current_tokens += doc_tokens
+
+            # 记录来源
+            sources.append({
+                "index": i + 1,
+                "doc_id": doc.doc_id,
+                "doc_name": doc.doc_name,
+                "section_title": doc.section_title,
+                "clause_number": doc.clause_number,
+                "page_number": doc.page_number,
+                "score": doc.score
+            })
+
+        context_text = "\n\n".join(context_parts)
+        return context_text, sources, truncated
+
+    def _format_single_doc(
+        self,
+        doc: RetrievedDocument,
+        index: int
+    ) -> str:
+        """格式化单个文档"""
+        parts = []
+
+        # 索引编号
+        parts.append(f"[{index}]")
+
+        # 元数据（可选）
+        if self.include_metadata:
+            meta_parts = []
+
+            if doc.doc_name:
+                meta_parts.append(f"文档: {doc.doc_name}")
+
+            if doc.section_title:
+                meta_parts.append(f"章节: {doc.section_title}")
+
+            if doc.clause_number:
+                clause_text = self.citation_format.format(clause=doc.clause_number)
+                meta_parts.append(clause_text)
+
+            if meta_parts:
+                parts.append(" | ".join(meta_parts))
+
+        # 内容
+        parts.append(doc.content)
+
+        return "\n".join(parts)
+
+    def _default_system_prompt(self) -> str:
+        """默认系统提示词"""
+        return """你是合规专家助手，基于提供的法规条款回答问题。
+
+回答要求：
+1. 直接回答问题，必须引用具体条款编号（如【条款5.2.1】）
+2. 如引用的条款不完整，说明需要进一步查阅原文
+3. 给出明确的合规建议和操作指导
+4. 如果检索内容不足以回答问题，如实说明
+
+回答格式：
+- 先给出直接结论
+- 然后引用支撑条款
+- 最后给出合规建议"""
+
+    def _estimate_tokens(self, text: str) -> int:
+        """估算文本token数"""
+        # 中文字符约1.5 token，英文约0.25 token
+        chinese_chars = sum(1 for c in text if '一' <= c <= '鿿')
+        other_chars = len(text) - chinese_chars
+        return int(chinese_chars * 1.5 + other_chars * 0.25)
+
+    def build_messages(
+        self,
+        context: RAGContext
+    ) -> List[Dict[str, str]]:
+        """
+        构建LLM消息格式
+
+        Args:
+            context: RAG上下文对象
+
+        Returns:
+            List[Dict]: [{"role": "system/user/assistant", "content": "..."}]
+        """
+        messages = [
+            {"role": "system", "content": context.system_prompt},
+            {"role": "user", "content": f"参考以下法规条款回答问题。\n\n{context.context_text}\n\n问题：{context.user_query}"}
+        ]
+
+        return messages
+
+
+def build_rag_context(
+    query: str,
+    documents: List[RetrievedDocument],
+    **kwargs
+) -> RAGContext:
+    """便捷函数：构建RAG上下文"""
+    builder = ContextBuilder()
+    return builder.build(query, documents, **kwargs)

backend/app/services/rag/prompt_templates.py

@@ -0,0 +1,296 @@
+# src/services/rag/prompt_templates.py
+"""RAG Prompt模板 - 合规问答专用Prompt"""
+
+from typing import Dict, Optional
+from dataclasses import dataclass
+
+
+@dataclass
+class PromptTemplate:
+    """Prompt模板"""
+    name: str
+    system_prompt: str
+    user_template: str
+    description: str
+
+
+class PromptTemplates:
+    """
+    合规问答Prompt模板库
+
+    包含多种场景的Prompt模板：
+    - 合规问答（标准）
+    - 条款解读（详细解释）
+    - 合规检查（判断合规状态）
+    - 差异对比（新旧法规对比）
+    - 报告生成（合规报告）
+    """
+
+    # 合规问答标准模板
+    COMPLIANCE_QA = PromptTemplate(
+        name="compliance_qa",
+        system_prompt="""你是合规专家助手，专门解答法规合规问题。
+
+角色定位：
+- 深谙国家法规标准（GB标准、行业标准）
+- 熟悉车辆安全、数据安全、EHS等领域合规要求
+- 提供专业、准确、可操作的合规建议
+
+回答规范：
+1. 必须引用具体条款编号（如【条款5.2.1】）
+2. 优先引用高相关性条款（score ≥ 0.5）
+3. 如条款内容不完整，明确提示需要查阅原文
+4. 给出明确的合规结论和建议
+5. 如检索内容不足以回答，如实说明
+
+回答格式：
+【结论】直接给出合规判断或答案
+
+【条款依据】
+- 【条款X.X.X】简要内容摘要（相关性: 高/中）
+- ...
+
+【合规建议】
+1. 具体操作建议
+2. 需要注意的风险点
+3. 后续行动建议""",
+        user_template="""请根据以下法规条款回答问题。
+
+【法规条款】
+{context}
+
+【用户问题】
+{query}""",
+        description="标准合规问答模板"
+    )
+
+    # 条款解读模板（详细解释）
+    CLAUSE_INTERPRETATION = PromptTemplate(
+        name="clause_interpretation",
+        system_prompt="""你是法规解读专家，负责详细解释法规条款的含义和应用。
+
+解读要求：
+1. 逐句解释条款原文的含义
+2. 说明条款的目的和背景
+3. 举例说明条款的实际应用场景
+4. 解释常见的误解和注意事项
+
+解读格式：
+【条款原文】完整引用条款
+
+【逐句解读】
+- "原文句1"：解读含义
+- "原文句2"：解读含义
+...
+
+【应用场景】
+具体举例说明该条款在实际工作中如何应用
+
+【注意事项】
+常见误解、执行难点、合规风险点""",
+        user_template="""请解读以下法规条款：
+
+条款编号：{clause_number}
+条款内容：{content}
+
+用户关注点：{query}""",
+        description="条款详细解读模板"
+    )
+
+    # 合规检查模板（判断合规状态）
+    COMPLIANCE_CHECK = PromptTemplate(
+        name="compliance_check",
+        system_prompt="""你是合规检查专家，负责评估企业行为或产品的合规状态。
+
+检查流程：
+1. 理解企业行为/产品描述
+2. 识别相关的法规条款
+3. 逐条对照检查合规状态
+4. 给出综合合规结论和整改建议
+
+合规状态分类：
+- ✅ 符合：完全满足法规要求
+- ⚠️ 需评估：需要进一步核实或补充材料
+- ❌ 不符合：明确违反法规要求
+- ❓ 无适用条款：检索内容不足以判断
+
+检查格式：
+【合规检查报告】
+
+一、检查对象
+{描述企业行为/产品}
+
+二、条款对照检查
+| 条款编号 | 要求摘要 | 检查状态 | 说明 |
+|--------|---------|---------|------|
+| 【条款X.X.X】 | ... | ✅/⚠️/❌/❓ | ... |
+
+三、综合结论
+合规等级：A/B/C/D（完全合规/基本合规/部分合规/不合规）
+
+四、整改建议（如需要）
+1. ...
+2. ...""",
+        user_template="""请对以下企业行为进行合规检查：
+
+【行为/产品描述】
+{query}
+
+【相关法规条款】
+{context}""",
+        description="合规检查评估模板"
+    )
+
+    # 差异对比模板（新旧法规对比）
+    COMPARISON = PromptTemplate(
+        name="comparison",
+        system_prompt="""你是法规变更分析专家，负责对比新旧法规版本的差异。
+
+对比任务：
+1. 识别新旧版本的条款差异
+2. 分类差异类型（新增/修改/删除）
+3. 分析差异的影响范围
+4. 给出企业应对建议
+
+差异分类：
+- 🆕 新增条款：原版本不存在
+- 🔄 修改条款：内容有实质性变更
+- ❌ 删除条款：原条款被移除
+- ⚖️ 调整条款：仅格式/编号调整，实质内容不变
+
+对比格式：
+【法规变更对比分析】
+
+一、变更概述
+- 旧版本：{version_old}
+- 新版本：{version_new}
+- 变更条款数：{count}
+
+二、差异明细
+| 类型 | 条款编号 | 旧版本内容 | 新版本内容 | 变化要点 |
+|-----|---------|-----------|-----------|---------|
+| 🆕 | X.X.X | - | ... | 新增要求... |
+
+三、影响分析
+- 高影响：{条款列表}
+- 中影响：{条款列表}
+- 低影响：{条款列表}
+
+四、应对建议
+1. 立即整改项
+2. 逐步调整项
+3. 持续关注项""",
+        user_template="""请对比分析以下法规差异：
+
+【用户问题】
+{query}
+
+【旧版本条款】
+{context_old}
+
+【新版本条款】
+{context_new}""",
+        description="法规版本对比模板"
+    )
+
+    # 报告生成模板
+    REPORT_GENERATION = PromptTemplate(
+        name="report_generation",
+        system_prompt="""你是合规报告撰写专家，负责生成结构化的合规分析报告。
+
+报告要求：
+1. 结构清晰、逻辑严谨
+2. 数据准确、引用规范
+3. 结论明确、建议可操作
+4. 语言专业、表达简洁
+
+报告结构：
+1. 概述（背景、范围）
+2. 分析内容（主体分析）
+3. 发现问题（合规差距）
+4. 整改建议（具体措施）
+5. 附录（引用条款原文）""",
+        user_template="""请生成以下合规报告：
+
+【报告主题】
+{query}
+
+【分析依据】
+{context}
+
+【报告要求】
+{requirements}""",
+        description="合规报告生成模板"
+    )
+
+    # 文档摘要生成模板
+    DOCUMENT_SUMMARY = PromptTemplate(
+        name="document_summary",
+        system_prompt="""你是法规文档摘要专家，负责生成法规文档的核心要点摘要。
+
+摘要要求：
+1. 精炼核心内容，不超过1024字
+2. 突出关键合规要求和条款编号
+3. 说明适用范围和生效条件
+4. 列出重要定义和术语解释
+
+摘要格式：
+【法规名称】{doc_name}
+
+【适用范围】{适用范围描述}
+
+【核心条款摘要】
+- 【条款X.X.X】{关键要求}（重要度：高）
+- ...
+
+【关键术语】
+- 术语1：定义解释
+- ...
+
+【合规要点】
+1. 必须满足的核心要求
+2. 需要特别注意的条款""",
+        user_template="""请生成以下法规文档的摘要：
+
+【文档名称】
+{doc_name}
+
+【文档内容】
+{content}
+
+请生成不超过1024字的摘要。""",
+        description="文档摘要生成模板"
+    )
+
+    @classmethod
+    def get_template(cls, name: str) -> Optional[PromptTemplate]:
+        """获取指定模板"""
+        templates = {
+            "compliance_qa": cls.COMPLIANCE_QA,
+            "clause_interpretation": cls.CLAUSE_INTERPRETATION,
+            "compliance_check": cls.COMPLIANCE_CHECK,
+            "comparison": cls.COMPARISON,
+            "report": cls.REPORT_GENERATION,
+            "document_summary": cls.DOCUMENT_SUMMARY
+        }
+        return templates.get(name)
+
+    @classmethod
+    def list_templates(cls) -> Dict[str, str]:
+        """列出所有模板"""
+        return {
+            "compliance_qa": cls.COMPLIANCE_QA.description,
+            "clause_interpretation": cls.CLAUSE_INTERPRETATION.description,
+            "compliance_check": cls.COMPLIANCE_CHECK.description,
+            "comparison": cls.COMPARISON.description,
+            "report": cls.REPORT_GENERATION.description,
+            "document_summary": cls.DOCUMENT_SUMMARY.description
+        }
+
+
+def get_prompt_template(name: str) -> PromptTemplate:
+    """便捷函数：获取Prompt模板"""
+    template = PromptTemplates.get_template(name)
+    if not template:
+        raise ValueError(f"不存在的模板: {name}")
+    return template

backend/app/services/rag/retriever.py

@@ -0,0 +1,193 @@
+# src/services/rag/retriever.py
+"""RAG检索服务 - 封装Milvus检索"""
+
+from typing import List, Dict, Optional, Any
+from dataclasses import dataclass, field
+from loguru import logger
+
+from app.services.embedding.bge_m3_embedder import BGEM3Embedder
+from app.services.storage.milvus_client import MilvusClient, SearchResult
+from app.config.settings import settings
+
+
+@dataclass
+class RetrievedDocument:
+    """检索到的文档"""
+    content: str
+    doc_id: str  # 文档ID，用于下载
+    doc_name: str
+    section_title: str
+    clause_number: str
+    page_number: int
+    score: float
+    metadata: Dict[str, Any] = field(default_factory=dict)
+
+
+class Retriever:
+    """
+    RAG检索器
+
+    功能：
+    - 向量检索（Dense + Sparse混合）
+    - 重排序（可选）
+    - 过滤和筛选
+    """
+
+    def __init__(
+        self,
+        top_k: int = None,
+        rerank: bool = False,
+        min_score: float = 0.3
+    ):
+        """
+        初始化检索器
+
+        Args:
+            top_k: 检索召回数量
+            rerank: 是否启用重排序
+            min_score: 最低相关性分数阈值
+        """
+        self.top_k = top_k or settings.rag_top_k
+        self.rerank = rerank
+        self.min_score = min_score
+
+        # 嵌入模型（延迟加载）
+        self.embedder: Optional[BGEM3Embedder] = None
+
+        # Milvus客户端（延迟连接）
+        self.milvus: Optional[MilvusClient] = None
+
+        logger.info(f"检索器初始化: top_k={self.top_k}, rerank={self.rerank}")
+
+    def _init_embedder(self):
+        """延迟初始化嵌入模型"""
+        if self.embedder is None:
+            logger.info("加载嵌入模型...")
+            self.embedder = BGEM3Embedder(model_name=settings.embedding_model)
+
+    def _init_milvus(self):
+        """延迟初始化Milvus"""
+        if self.milvus is None:
+            logger.info("连接Milvus...")
+            self.milvus = MilvusClient()
+            self.milvus.connect()
+            self.milvus.create_collection(recreate=False)
+            self.milvus.load_collection()
+
+    def retrieve(
+        self,
+        query: str,
+        filters: Optional[str] = None,
+        top_k: Optional[int] = None
+    ) -> List[RetrievedDocument]:
+        """
+        检索相关文档
+
+        Args:
+            query: 查询文本
+            filters: 过滤条件（如 "regulation_type=='车辆安全'"）
+            top_k: 返回数量（可选，覆盖默认值）
+
+        Returns:
+            List[RetrievedDocument]: 检索结果列表
+        """
+        logger.info(f"执行检索: {query}")
+
+        # 初始化组件
+        self._init_embedder()
+        self._init_milvus()
+
+        # 生成查询向量
+        query_embedding = self.embedder.embed_single(query)
+
+        # 执行混合检索
+        results = self.milvus.hybrid_search(
+            query_dense=query_embedding['dense'].tolist(),
+            query_sparse=query_embedding['sparse'],
+            top_k=top_k or self.top_k,
+            filters=filters
+        )
+
+        # 转换为RetrievedDocument格式
+        documents = []
+        for r in results:
+            if r.score >= self.min_score:
+                doc = RetrievedDocument(
+                    content=r.content,
+                    doc_id=r.metadata.get("doc_id", ""),
+                    doc_name=r.metadata.get("doc_name", ""),
+                    section_title=r.metadata.get("section_title", ""),
+                    clause_number=r.metadata.get("clause_number", ""),
+                    page_number=r.metadata.get("page_number", 0),
+                    score=r.score,
+                    metadata=r.metadata
+                )
+                documents.append(doc)
+
+        logger.success(f"检索完成，返回{len(documents)}条结果（阈值过滤后）")
+        return documents
+
+    def retrieve_with_scores(
+        self,
+        query: str,
+        filters: Optional[str] = None
+    ) -> List[Dict]:
+        """
+        检索并返回完整结果（包含分数）
+
+        Args:
+            query: 查询文本
+            filters: 过滤条件
+
+        Returns:
+            List[Dict]: 包含分数的检索结果
+        """
+        documents = self.retrieve(query, filters)
+        return [
+            {
+                "content": doc.content,
+                "doc_id": doc.doc_id,
+                "doc_name": doc.doc_name,
+                "section_title": doc.section_title,
+                "clause_number": doc.clause_number,
+                "page_number": doc.page_number,
+                "score": doc.score
+            }
+            for doc in documents
+        ]
+
+    def search_by_doc_name(
+        self,
+        query: str,
+        doc_name: str
+    ) -> List[RetrievedDocument]:
+        """按文档名称过滤检索"""
+        filters = f'doc_name=="{doc_name}"'
+        return self.retrieve(query, filters)
+
+    def search_by_regulation_type(
+        self,
+        query: str,
+        regulation_type: str
+    ) -> List[RetrievedDocument]:
+        """按法规类型过滤检索"""
+        filters = f'regulation_type=="{regulation_type}"'
+        return self.retrieve(query, filters)
+
+    def close(self):
+        """关闭连接"""
+        if self.milvus:
+            self.milvus.disconnect()
+        logger.info("检索器已关闭")
+
+
+def retrieve_regulations(
+    query: str,
+    top_k: int = 10,
+    filters: Optional[str] = None
+) -> List[RetrievedDocument]:
+    """便捷函数：检索法规"""
+    retriever = Retriever(top_k=top_k)
+    results = retriever.retrieve(query, filters)
+    retriever.close()
+    return results

backend/app/services/storage/__init__.py

@@ -0,0 +1,7 @@
+# src/services/storage/__init__.py
+"""存储服务"""
+
+from .milvus_client import MilvusClient
+from .minio_client import MinIOClient
+
+__all__ = ["MilvusClient", "MinIOClient"]

backend/app/services/storage/milvus_client.py

@@ -0,0 +1,485 @@
+# src/services/storage/milvus_client.py
+"""Milvus向量数据库客户端 - 存储与检索服务"""
+
+from pymilvus import (
+    connections,
+    Collection,
+    FieldSchema,
+    CollectionSchema,
+    DataType,
+    utility
+)
+from typing import List, Dict, Optional, Any
+from dataclasses import dataclass, field
+from loguru import logger
+import time
+import numpy as np
+
+from ..embedding.text_chunker import TextChunk
+from ..embedding.bge_m3_embedder import EmbeddingResult
+from app.config.settings import settings
+
+
+@dataclass
+class SearchResult:
+    """检索结果"""
+    id: int
+    content: str
+    score: float
+    metadata: Dict[str, Any] = field(default_factory=dict)
+
+
+@dataclass
+class MilvusDocument:
+    """Milvus文档数据结构"""
+    doc_id: str
+    chunk_id: str
+    content: str
+    dense_vector: List[float]
+    sparse_vector: Dict[int, float]
+    doc_name: str
+    section_title: str
+    clause_number: str
+    page_number: int
+    regulation_type: str
+    version: str
+    create_time: int
+
+
+class MilvusClient:
+    """Milvus向量数据库客户端"""
+
+    COLLECTION_NAME = "regulations"
+
+    SCHEMA_FIELDS = [
+        FieldSchema(name="id", dtype=DataType.INT64, is_primary=True, auto_id=True),
+        FieldSchema(name="doc_id", dtype=DataType.VARCHAR, max_length=64),
+        FieldSchema(name="chunk_id", dtype=DataType.VARCHAR, max_length=128),
+        FieldSchema(name="content", dtype=DataType.VARCHAR, max_length=8192),
+        FieldSchema(name="dense_vector", dtype=DataType.FLOAT_VECTOR, dim=1024),
+        FieldSchema(name="sparse_vector", dtype=DataType.SPARSE_FLOAT_VECTOR),
+        FieldSchema(name="doc_name", dtype=DataType.VARCHAR, max_length=256),
+        FieldSchema(name="section_title", dtype=DataType.VARCHAR, max_length=512),
+        FieldSchema(name="clause_number", dtype=DataType.VARCHAR, max_length=64),
+        FieldSchema(name="page_number", dtype=DataType.INT64),
+        FieldSchema(name="regulation_type", dtype=DataType.VARCHAR, max_length=32),
+        FieldSchema(name="version", dtype=DataType.VARCHAR, max_length=32),
+        FieldSchema(name="create_time", dtype=DataType.INT64),
+    ]
+
+    def __init__(
+        self,
+        host: str = None,
+        port: int = None,
+        collection_name: str = None,
+        db_name: str = None
+    ):
+        self.host = host or settings.milvus_host
+        self.port = port or settings.milvus_port
+        self.collection_name = collection_name or settings.milvus_collection
+        self.db_name = db_name or settings.milvus_db_name
+
+        self.collection: Optional[Collection] = None
+        self.connected = False
+
+        logger.info(f"Milvus客户端配置: {self.host}:{self.port}, Collection: {self.collection_name}")
+
+    def connect(self) -> bool:
+        """连接到Milvus服务器"""
+        try:
+            connections.connect(
+                alias="default",
+                host=self.host,
+                port=self.port,
+                db_name=self.db_name
+            )
+            self.connected = True
+            logger.success(f"Milvus连接成功: {self.host}:{self.port}")
+            return True
+        except Exception as e:
+            logger.error(f"Milvus连接失败: {e}")
+            self.connected = False
+            return False
+
+    def disconnect(self):
+        """断开连接"""
+        try:
+            connections.disconnect("default")
+            self.connected = False
+            logger.info("Milvus连接已断开")
+        except Exception as e:
+            logger.warning(f"断开连接时出错: {e}")
+
+    def create_collection(self, recreate: bool = False) -> bool:
+        """创建Collection"""
+        if not self.connected:
+            logger.warning("未连接到Milvus，请先调用connect()")
+            return False
+
+        try:
+            if utility.has_collection(self.collection_name):
+                if recreate:
+                    logger.info(f"删除已存在的Collection: {self.collection_name}")
+                    utility.drop_collection(self.collection_name)
+                else:
+                    logger.info(f"Collection已存在: {self.collection_name}")
+                    self.collection = Collection(self.collection_name)
+                    return True
+
+            schema = CollectionSchema(
+                fields=self.SCHEMA_FIELDS,
+                description="法规文档向量存储",
+                enable_dynamic_field=True
+            )
+
+            self.collection = Collection(
+                name=self.collection_name,
+                schema=schema
+            )
+
+            self._create_indexes()
+
+            logger.success(f"Collection创建成功: {self.collection_name}")
+            return True
+
+        except Exception as e:
+            logger.error(f"Collection创建失败: {e}")
+            return False
+
+    def _create_indexes(self):
+        """创建向量索引"""
+        if not self.collection:
+            return
+
+        try:
+            dense_index_params = {
+                "metric_type": "COSINE",
+                "index_type": "IVF_FLAT",
+                "params": {"nlist": 128}
+            }
+            self.collection.create_index(
+                field_name="dense_vector",
+                index_params=dense_index_params
+            )
+
+            sparse_index_params = {
+                "metric_type": "IP",
+                "index_type": "SPARSE_INVERTED_INDEX",
+                "params": {"drop_ratio_build": 0.2}
+            }
+            self.collection.create_index(
+                field_name="sparse_vector",
+                index_params=sparse_index_params
+            )
+
+            logger.success("向量索引创建成功")
+
+        except Exception as e:
+            logger.warning(f"创建索引时出错: {e}")
+
+    def load_collection(self):
+        """加载Collection到内存"""
+        if self.collection:
+            self.collection.load()
+            logger.info(f"Collection已加载: {self.collection_name}")
+
+    def release_collection(self):
+        """释放Collection内存"""
+        if self.collection:
+            self.collection.release()
+            logger.info(f"Collection已释放: {self.collection_name}")
+
+    def insert_chunks(
+        self,
+        chunks: List[TextChunk],
+        embeddings: EmbeddingResult
+    ) -> List[int]:
+        """插入文档分块和嵌入向量"""
+        if not self.collection:
+            logger.warning("Collection未初始化")
+            return []
+
+        if len(chunks) != len(embeddings.texts):
+            logger.warning(f"Chunks数量与嵌入数量不匹配")
+            return []
+
+        logger.info(f"准备插入{len(chunks)}个文档分块")
+
+        try:
+            data = []
+            current_time = int(time.time())
+
+            for chunk, dense_emb, sparse_emb in zip(
+                chunks,
+                embeddings.dense_embeddings,
+                embeddings.sparse_embeddings
+            ):
+                row = {
+                    "doc_id": chunk.metadata.doc_id,
+                    "chunk_id": chunk.metadata.chunk_id,
+                    "content": chunk.content,
+                    "dense_vector": dense_emb.tolist(),
+                    "sparse_vector": sparse_emb,
+                    "doc_name": chunk.metadata.doc_name,
+                    "section_title": chunk.metadata.section_title,
+                    "clause_number": chunk.metadata.clause_number,
+                    "page_number": chunk.metadata.page_number,
+                    "regulation_type": chunk.metadata.regulation_type,
+                    "version": chunk.metadata.version,
+                    "create_time": current_time
+                }
+                data.append(row)
+
+            result = self.collection.insert(data)
+            self.collection.flush()
+
+            logger.success(f"插入完成，共{len(result.primary_keys)}条记录")
+            return result.primary_keys
+
+        except Exception as e:
+            logger.error(f"插入数据失败: {e}")
+            return []
+
+    def hybrid_search(
+        self,
+        query_dense: List[float],
+        query_sparse: Dict[int, float],
+        top_k: int = 10,
+        filters: Optional[str] = None
+    ) -> List[SearchResult]:
+        """混合检索：Dense + Sparse"""
+        if not self.collection:
+            logger.warning("Collection未初始化")
+            return []
+
+        try:
+            self.collection.load()
+
+            # 使用简单的Dense检索（兼容所有版本）
+            dense_results = self.dense_search(query_dense, top_k, filters)
+
+            # 可选：合并Sparse结果
+            if query_sparse:
+                sparse_results = self.sparse_search(query_sparse, top_k, filters)
+                merged = self._merge_results(dense_results, sparse_results, top_k)
+                logger.success(f"混合检索完成，返回{len(merged)}条结果")
+                return merged
+
+            return dense_results
+
+        except Exception as e:
+            logger.error(f"混合检索失败: {e}")
+            return []
+
+    def _merge_results(
+        self,
+        dense_results: List[SearchResult],
+        sparse_results: List[SearchResult],
+        top_k: int,
+        dense_weight: float = 0.6
+    ) -> List[SearchResult]:
+        """手动融合Dense和Sparse结果"""
+        sparse_weight = 1 - dense_weight
+        merged_dict = {}
+
+        for r in dense_results:
+            merged_dict[r.id] = {
+                "result": r,
+                "dense_score": r.score * dense_weight,
+                "sparse_score": 0
+            }
+
+        for r in sparse_results:
+            if r.id in merged_dict:
+                merged_dict[r.id]["sparse_score"] = r.score * sparse_weight
+            else:
+                merged_dict[r.id] = {
+                    "result": r,
+                    "dense_score": 0,
+                    "sparse_score": r.score * sparse_weight
+                }
+
+        final_results = []
+        for id_, data in merged_dict.items():
+            result = data["result"]
+            final_score = data["dense_score"] + data["sparse_score"]
+            final_results.append(SearchResult(
+                id=result.id,
+                content=result.content,
+                score=final_score,
+                metadata=result.metadata
+            ))
+
+        final_results.sort(key=lambda x: x.score, reverse=True)
+        return final_results[:top_k]
+
+    def dense_search(
+        self,
+        query_dense: List[float],
+        top_k: int = 10,
+        filters: Optional[str] = None
+    ) -> List[SearchResult]:
+        """纯Dense向量检索"""
+        if not self.collection:
+            return []
+
+        try:
+            self.collection.load()
+
+            search_params = {
+                "metric_type": "COSINE",
+                "params": {"nprobe": 16}
+            }
+
+            results = self.collection.search(
+                data=[query_dense],
+                anns_field="dense_vector",
+                param=search_params,
+                limit=top_k,
+                filter=filters,
+                output_fields=[
+                    "doc_id", "chunk_id", "content",
+                    "doc_name", "section_title", "clause_number",
+                    "page_number", "regulation_type", "version"
+                ]
+            )
+
+            search_results = []
+            for hits in results:
+                for hit in hits:
+                    result = SearchResult(
+                        id=hit.id,
+                        content=hit.entity.get("content", ""),
+                        score=hit.score,
+                        metadata={
+                            "doc_id": hit.entity.get("doc_id", ""),
+                            "chunk_id": hit.entity.get("chunk_id", ""),
+                            "doc_name": hit.entity.get("doc_name", ""),
+                            "section_title": hit.entity.get("section_title", ""),
+                            "clause_number": hit.entity.get("clause_number", ""),
+                            "page_number": hit.entity.get("page_number", 0),
+                            "regulation_type": hit.entity.get("regulation_type", ""),
+                            "version": hit.entity.get("version", ""),
+                        }
+                    )
+                    search_results.append(result)
+
+            return search_results
+
+        except Exception as e:
+            logger.error(f"Dense检索失败: {e}")
+            return []
+
+    def sparse_search(
+        self,
+        query_sparse: Dict[int, float],
+        top_k: int = 10,
+        filters: Optional[str] = None
+    ) -> List[SearchResult]:
+        """纯Sparse向量检索"""
+        if not self.collection:
+            return []
+
+        try:
+            self.collection.load()
+
+            search_params = {
+                "metric_type": "IP",
+                "params": {"drop_ratio_search": 0.2}
+            }
+
+            results = self.collection.search(
+                data=[query_sparse],
+                anns_field="sparse_vector",
+                param=search_params,
+                limit=top_k,
+                filter=filters,
+                output_fields=[
+                    "doc_id", "chunk_id", "content",
+                    "doc_name", "section_title", "clause_number",
+                    "page_number", "regulation_type", "version"
+                ]
+            )
+
+            search_results = []
+            for hits in results:
+                for hit in hits:
+                    result = SearchResult(
+                        id=hit.id,
+                        content=hit.entity.get("content", ""),
+                        score=hit.score,
+                        metadata={
+                            "doc_id": hit.entity.get("doc_id", ""),
+                            "chunk_id": hit.entity.get("chunk_id", ""),
+                            "doc_name": hit.entity.get("doc_name", ""),
+                            "section_title": hit.entity.get("section_title", ""),
+                            "clause_number": hit.entity.get("clause_number", ""),
+                            "page_number": hit.entity.get("page_number", 0),
+                            "regulation_type": hit.entity.get("regulation_type", ""),
+                            "version": hit.entity.get("version", ""),
+                        }
+                    )
+                    search_results.append(result)
+
+            return search_results
+
+        except Exception as e:
+            logger.error(f"Sparse检索失败: {e}")
+            return []
+
+    def delete_by_doc_id(self, doc_id: str) -> int:
+        """根据doc_id删除记录"""
+        if not self.collection:
+            return 0
+
+        try:
+            expr = f'doc_id=="{doc_id}"'
+            result = self.collection.delete(expr)
+            logger.info(f"删除记录: doc_id={doc_id}, 数量={len(result.primary_keys)}")
+            return len(result.primary_keys)
+        except Exception as e:
+            logger.error(f"删除失败: {e}")
+            return 0
+
+    def get_collection_stats(self) -> Dict[str, Any]:
+        """获取Collection统计信息"""
+        if not self.collection:
+            return {}
+
+        try:
+            stats = {
+                "name": self.collection_name,
+                "num_entities": self.collection.num_entities,
+                "description": self.collection.description,
+            }
+            return stats
+        except Exception as e:
+            logger.warning(f"获取统计信息失败: {e}")
+            return {}
+
+
+def create_milvus_client() -> MilvusClient:
+    """便捷函数：创建Milvus客户端"""
+    client = MilvusClient()
+    client.connect()
+    client.create_collection(recreate=False)
+    return client
+
+
+def insert_documents(
+    client: MilvusClient,
+    chunks: List[TextChunk],
+    embeddings: EmbeddingResult
+) -> List[int]:
+    """便捷函数：插入文档"""
+    return client.insert_chunks(chunks, embeddings)
+
+
+def search_regulations(
+    client: MilvusClient,
+    query_dense: List[float],
+    query_sparse: Dict[int, float],
+    top_k: int = 10
+) -> List[SearchResult]:
+    """便捷函数：检索法规"""
+    return client.hybrid_search(query_dense, query_sparse, top_k)

backend/app/services/storage/minio_client.py

@@ -0,0 +1,352 @@
+# src/services/storage/minio_client.py
+"""MinIO对象存储客户端 - 文档文件存储"""
+
+from minio import Minio
+from minio.error import S3Error
+from typing import Optional, Dict, Any
+from loguru import logger
+from io import BytesIO
+import os
+
+from app.config.settings import settings
+
+
+class MinIOClient:
+    """MinIO对象存储客户端"""
+
+    def __init__(
+        self,
+        endpoint: str = None,
+        access_key: str = None,
+        secret_key: str = None,
+        bucket: str = None,
+        secure: bool = None
+    ):
+        """
+        初始化MinIO客户端
+
+        Args:
+            endpoint: MinIO服务地址
+            access_key: 访问密钥
+            secret_key: 秘密密钥
+            bucket: 存储桶名称
+            secure: 是否使用HTTPS
+        """
+        self.endpoint = endpoint or settings.minio_endpoint
+        self.access_key = access_key or settings.minio_access_key
+        self.secret_key = secret_key or settings.minio_secret_key
+        self.bucket = bucket or settings.minio_bucket
+        self.secure = secure or settings.minio_secure
+
+        self.client: Optional[Minio] = None
+        self.connected = False
+
+        logger.info(f"MinIO客户端配置: {self.endpoint}, bucket={self.bucket}")
+
+    def connect(self) -> bool:
+        """连接MinIO服务"""
+        try:
+            self.client = Minio(
+                self.endpoint,
+                access_key=self.access_key,
+                secret_key=self.secret_key,
+                secure=self.secure
+            )
+            self.connected = True
+            logger.success(f"MinIO连接成功: {self.endpoint}")
+            return True
+        except Exception as e:
+            logger.error(f"MinIO连接失败: {e}")
+            self.connected = False
+            return False
+
+    def ensure_bucket(self) -> bool:
+        """确保存储桶存在"""
+        if not self.connected:
+            logger.warning("未连接MinIO，请先调用connect()")
+            return False
+
+        try:
+            if not self.client.bucket_exists(self.bucket):
+                self.client.make_bucket(self.bucket)
+                logger.success(f"创建存储桶: {self.bucket}")
+            else:
+                logger.info(f"存储桶已存在: {self.bucket}")
+            return True
+        except S3Error as e:
+            logger.error(f"存储桶操作失败: {e}")
+            return False
+
+    def upload_file(
+        self,
+        file_path: str,
+        object_name: str,
+        metadata: Dict[str, Any] = None
+    ) -> bool:
+        """
+        上传本地文件到MinIO
+
+        Args:
+            file_path: 本地文件路径
+            object_name: MinIO对象名称
+            metadata: 元数据
+
+        Returns:
+            bool: 是否成功
+        """
+        if not self.connected:
+            self.connect()
+            self.ensure_bucket()
+
+        try:
+            file_size = os.stat(file_path).st_size
+            content_type = self._get_content_type(file_path)
+
+            with open(file_path, 'rb') as f:
+                self.client.put_object(
+                    self.bucket,
+                    object_name,
+                    f,
+                    file_size,
+                    content_type=content_type,
+                    metadata=metadata
+                )
+
+            logger.success(f"文件上传成功: {object_name}, 大小={file_size}")
+            return True
+
+        except S3Error as e:
+            logger.error(f"文件上传失败: {e}")
+            return False
+
+    def upload_bytes(
+        self,
+        data: bytes,
+        object_name: str,
+        content_type: str = "application/octet-stream",
+        metadata: Dict[str, Any] = None
+    ) -> bool:
+        """
+        上传字节数据到MinIO
+
+        Args:
+            data: 文件字节数据
+            object_name: MinIO对象名称
+            content_type: 内容类型
+            metadata: 元数据（注意：MinIO仅支持US-ASCII字符）
+
+        Returns:
+            bool: 是否成功
+        """
+        if not self.connected:
+            self.connect()
+            self.ensure_bucket()
+
+        try:
+            data_stream = BytesIO(data)
+
+            # 处理metadata：仅保留ASCII安全字符
+            safe_metadata = None
+            if metadata:
+                safe_metadata = {}
+                for key, value in metadata.items():
+                    if isinstance(value, str):
+                        # 只保留ASCII字符或转换为安全格式
+                        try:
+                            value.encode('ascii')
+                            safe_metadata[key] = value
+                        except UnicodeEncodeError:
+                            # 中文字符跳过或用占位符
+                            safe_metadata[key] = ""
+                    else:
+                        safe_metadata[key] = str(value)
+
+            self.client.put_object(
+                self.bucket,
+                object_name,
+                data_stream,
+                len(data),
+                content_type=content_type,
+                metadata=safe_metadata
+            )
+
+            logger.success(f"数据上传成功: {object_name}, 大小={len(data)}")
+            return True
+
+        except S3Error as e:
+            logger.error(f"数据上传失败: {e}")
+            return False
+
+    def download_file(
+        self,
+        object_name: str,
+        file_path: str
+    ) -> bool:
+        """
+        从MinIO下载文件到本地
+
+        Args:
+            object_name: MinIO对象名称
+            file_path: 本地保存路径
+
+        Returns:
+            bool: 是否成功
+        """
+        if not self.connected:
+            self.connect()
+
+        try:
+            self.client.fget_object(
+                self.bucket,
+                object_name,
+                file_path
+            )
+            logger.success(f"文件下载成功: {object_name} -> {file_path}")
+            return True
+
+        except S3Error as e:
+            logger.error(f"文件下载失败: {e}")
+            return False
+
+    def get_object_url(
+        self,
+        object_name: str,
+        expires: int = 3600
+    ) -> Optional[str]:
+        """
+        获取对象下载URL（临时URL）
+
+        Args:
+            object_name: MinIO对象名称
+            expires: URL有效期（秒）
+
+        Returns:
+            str: 下载URL
+        """
+        if not self.connected:
+            self.connect()
+
+        try:
+            url = self.client.presigned_get_object(
+                self.bucket,
+                object_name,
+                expires=expires
+            )
+            return url
+
+        except S3Error as e:
+            logger.error(f"获取URL失败: {e}")
+            return None
+
+    def get_object_data(self, object_name: str) -> Optional[bytes]:
+        """
+        获取对象数据（字节）
+
+        Args:
+            object_name: MinIO对象名称
+
+        Returns:
+            bytes: 文件数据
+        """
+        if not self.connected:
+            self.connect()
+
+        try:
+            response = self.client.get_object(self.bucket, object_name)
+            data = response.read()
+            response.close()
+            response.release_conn()
+            return data
+
+        except S3Error as e:
+            logger.error(f"获取对象数据失败: {e}")
+            return None
+
+    def delete_object(self, object_name: str) -> bool:
+        """
+        删除对象
+
+        Args:
+            object_name: MinIO对象名称
+
+        Returns:
+            bool: 是否成功
+        """
+        if not self.connected:
+            self.connect()
+
+        try:
+            self.client.remove_object(self.bucket, object_name)
+            logger.info(f"对象删除成功: {object_name}")
+            return True
+
+        except S3Error as e:
+            logger.error(f"对象删除失败: {e}")
+            return False
+
+    def list_objects(self, prefix: str = "") -> list:
+        """
+        列出存储桶中的对象
+
+        Args:
+            prefix: 对象名称前缀
+
+        Returns:
+            list: 对象列表
+        """
+        if not self.connected:
+            self.connect()
+
+        try:
+            objects = self.client.list_objects(self.bucket, prefix=prefix)
+            return [obj.object_name for obj in objects]
+
+        except S3Error as e:
+            logger.error(f"列出对象失败: {e}")
+            return []
+
+    def object_exists(self, object_name: str) -> bool:
+        """
+        检查对象是否存在
+
+        Args:
+            object_name: MinIO对象名称
+
+        Returns:
+            bool: 是否存在
+        """
+        if not self.connected:
+            self.connect()
+
+        try:
+            self.client.stat_object(self.bucket, object_name)
+            return True
+
+        except S3Error:
+            return False
+
+    def _get_content_type(self, file_path: str) -> str:
+        """根据文件扩展名获取Content-Type"""
+        ext = os.path.splitext(file_path)[1].lower()
+        content_types = {
+            '.pdf': 'application/pdf',
+            '.doc': 'application/msword',
+            '.docx': 'application/vnd.openxmlformats-officedocument.wordprocessingml.document',
+            '.txt': 'text/plain',
+            '.json': 'application/json',
+            '.xml': 'application/xml',
+        }
+        return content_types.get(ext, 'application/octet-stream')
+
+    def close(self):
+        """关闭连接（MinIO客户端无需显式关闭）"""
+        self.connected = False
+        logger.info("MinIO客户端已关闭")
+
+
+def create_minio_client() -> MinIOClient:
+    """便捷函数：创建MinIO客户端"""
+    client = MinIOClient()
+    client.connect()
+    client.ensure_bucket()
+    return client

backend/app/utils/__init__.py

@@ -0,0 +1,4 @@
+from .chunking import TextChunker, chunker
+from .logger import logger, setup_logging
+
+__all__ = ["TextChunker", "chunker", "logger", "setup_logging"]

backend/app/utils/chunking.py

@@ -0,0 +1,78 @@
+import re
+from typing import List
+from app.core.config import settings
+
+
+class TextChunker:
+    def __init__(
+        self,
+        chunk_size: int = settings.chunk_size,
+        chunk_overlap: int = settings.chunk_overlap,
+    ):
+        self.chunk_size = chunk_size
+        self.chunk_overlap = chunk_overlap
+
+    def chunk_by_clause(self, text: str) -> List[dict]:
+        """按条款边界分块（适用于法规文档）"""
+        clause_pattern = r"(第[一二三四五六七八九十百]+条)"
+        parts = re.split(clause_pattern, text)
+
+        chunks = []
+        current_clause = None
+        current_text = ""
+        chunk_index = 0
+
+        for part in parts:
+            if re.match(clause_pattern, part):
+                if current_clause and current_text.strip():
+                    chunks.append({
+                        "clause_id": current_clause,
+                        "content": current_text.strip(),
+                        "chunk_index": chunk_index,
+                    })
+                    chunk_index += 1
+                current_clause = part
+                current_text = ""
+            else:
+                current_text += part
+
+        if current_clause and current_text.strip():
+            chunks.append({
+                "clause_id": current_clause,
+                "content": current_text.strip(),
+                "chunk_index": chunk_index,
+            })
+
+        return chunks
+
+    def chunk_by_size(self, text: str) -> List[dict]:
+        """按固定大小分块"""
+        chunks = []
+        start = 0
+        chunk_index = 0
+
+        while start < len(text):
+            end = start + self.chunk_size
+            chunk_text = text[start:end]
+
+            if chunk_text.strip():
+                chunks.append({
+                    "content": chunk_text.strip(),
+                    "chunk_index": chunk_index,
+                    "start_pos": start,
+                    "end_pos": end,
+                })
+                chunk_index += 1
+
+            start = end - self.chunk_overlap
+
+        return chunks
+
+    def estimate_tokens(self, text: str) -> int:
+        """估算token数量"""
+        chinese_chars = len(re.findall(r"[^\x00-\xff]", text))
+        english_chars = len(text) - chinese_chars
+        return int(chinese_chars / 1.5 + english_chars / 4)
+
+
+chunker = TextChunker()

backend/app/utils/logger.py

@@ -0,0 +1,24 @@
+import logging
+import sys
+
+
+def setup_logging() -> logging.Logger:
+    """配置日志"""
+    logger = logging.getLogger("app")
+    logger.setLevel(logging.INFO)
+
+    handler = logging.StreamHandler(sys.stdout)
+    handler.setLevel(logging.INFO)
+
+    formatter = logging.Formatter(
+        fmt="%(asctime)s | %(levelname)s | %(name)s | %(message)s",
+        datefmt="%Y-%m-%d %H:%M:%S",
+    )
+    handler.setFormatter(formatter)
+
+    logger.addHandler(handler)
+
+    return logger
+
+
+logger = setup_logging()

backend/app/workers/__init__.py

@@ -0,0 +1,2 @@
+# src/workers/__init__.py
+"""异步任务Worker模块"""

backend/app/workflows/__init__.py

@@ -0,0 +1,12 @@
+from .rag_workflow import RagState, rag_workflow, run_rag_workflow, stream_rag_workflow
+from .compliance_workflow import ComplianceState, compliance_workflow, run_compliance_workflow
+
+__all__ = [
+    "RagState",
+    "rag_workflow",
+    "run_rag_workflow",
+    "stream_rag_workflow",
+    "ComplianceState",
+    "compliance_workflow",
+    "run_compliance_workflow",
+]

backend/app/workflows/compliance_workflow.py

@@ -0,0 +1,175 @@
+from typing import TypedDict, List
+from langgraph.graph import StateGraph, END
+
+
+class ComplianceState(TypedDict):
+    document_path: str
+    raw_text: str
+    segments: List[dict]
+    matched_regulations: List[dict]
+    risk_dashboard: dict
+    priority_actions: List[dict]
+
+
+def parse_document(state: ComplianceState) -> dict:
+    """解析文档"""
+    from app.services import get_document_service
+    doc_service = get_document_service(
+        "/airegulation/demo-mao/backend/data/raw",
+        "/airegulation/demo-mao/backend/data/parsed",
+    )
+    text = doc_service.parse_document(state["document_path"])
+    return {"raw_text": text}
+
+
+def segment_document(state: ComplianceState) -> dict:
+    """AI语义分段"""
+    from app.services import llm_service
+    prompt = f"""请分析以下设计方案文档，按照设计意图将其分成若干语义段落。
+
+文档内容：
+{state['raw_text'][:3000]}
+
+请输出JSON格式的分段结果，每个段落包含：
+- intent: 段落意图/主题
+- startPos: 在原文中的起始位置（大致）
+- endPos: 在原文中的结束位置（大致）
+- keywords: 关键词列表
+
+输出格式：
+[{{"intent": "...", "startPos": 0, "endPos": 100, "keywords": [...]}}]"""
+
+    # 简化处理：返回基本分段
+    segments = [
+        {
+            "id": 1,
+            "intent": "整体设计概述",
+            "content": state["raw_text"][:500],
+            "keywords": ["设计", "方案"],
+        }
+    ]
+
+    return {"segments": segments}
+
+
+def match_regulations(state: ComplianceState) -> dict:
+    """法规匹配"""
+    from app.services import embedding_service, milvus_service
+    matched = []
+
+    for segment in state["segments"]:
+        keyword_text = " ".join(segment.get("keywords", []))
+        embedding = embedding_service.embed_single(keyword_text)
+
+        docs = milvus_service.search(embedding, top_k=5)
+
+        segment_regs = []
+        for doc in docs:
+            category = "high" if doc["score"] > 0.85 else ("medium" if doc["score"] > 0.6 else "low")
+            segment_regs.append({
+                "id": doc["id"],
+                "name": doc["doc_name"],
+                "clause": doc.get("clause_id"),
+                "score": doc["score"],
+                "match_keyword": keyword_text,
+                "category": category,
+                "full_content": doc["content"],
+            })
+
+        segment["regulations"] = segment_regs
+        matched.append(segment)
+
+    return {"matched_regulations": matched}
+
+
+def calculate_risk(state: ComplianceState) -> dict:
+    """计算风险等级"""
+    segments = state["matched_regulations"]
+
+    high_count = 0
+    medium_count = 0
+    low_count = 0
+    need_fix = 0
+    total_score = 0
+
+    for segment in segments:
+        regs = segment.get("regulations", [])
+        high_regs = [r for r in regs if r["category"] == "high"]
+
+        if high_regs:
+            avg_score = sum(r["score"] for r in high_regs) / len(high_regs)
+            if avg_score < 0.9:
+                segment["risk_level"] = "high"
+                high_count += 1
+                need_fix += 1
+            elif avg_score < 0.92:
+                segment["risk_level"] = "medium"
+                medium_count += 1
+            else:
+                segment["risk_level"] = "low"
+                low_count += 1
+        else:
+            segment["risk_level"] = "low"
+            low_count += 1
+
+        total_score += avg_score if high_regs else 100
+
+    avg_score = total_score / len(segments) if segments else 100
+
+    status = "pass" if avg_score >= 90 else ("warning" if avg_score >= 70 else "fail")
+    status_label = "合规" if status == "pass" else ("需要修改" if status == "warning" else "高风险")
+
+    dashboard = {
+        "score": avg_score,
+        "high_risk_count": high_count,
+        "medium_risk_count": medium_count,
+        "low_risk_count": low_count,
+        "need_fix_segments": need_fix,
+        "status": status,
+        "status_label": status_label,
+    }
+
+    return {"risk_dashboard": dashboard, "segments": segments}
+
+
+def generate_suggestions(state: ComplianceState) -> dict:
+    """生成优先建议"""
+    actions = []
+
+    for segment in state["segments"]:
+        for reg in segment.get("regulations", []):
+            if reg["category"] == "high" and reg["score"] < 0.9:
+                actions.append({
+                    "regulation": reg["name"],
+                    "issue": reg["match_keyword"],
+                    "suggestion": f"建议对照{reg['name']}第{reg.get('clause', '')}条进行修改",
+                    "severity": "high",
+                })
+
+    return {"priority_actions": actions}
+
+
+# 构建工作流图
+compliance_graph = StateGraph(ComplianceState)
+
+compliance_graph.add_node("parse", parse_document)
+compliance_graph.add_node("segment", segment_document)
+compliance_graph.add_node("match", match_regulations)
+compliance_graph.add_node("score", calculate_risk)
+compliance_graph.add_node("suggest", generate_suggestions)
+
+compliance_graph.set_entry_point("parse")
+compliance_graph.add_edge("parse", "segment")
+compliance_graph.add_edge("segment", "match")
+compliance_graph.add_edge("match", "score")
+compliance_graph.add_edge("score", "suggest")
+compliance_graph.add_edge("suggest", END)
+
+compliance_workflow = compliance_graph.compile()
+
+
+async def run_compliance_workflow(document_path: str) -> ComplianceState:
+    """运行合规分析工作流"""
+    initial_state: ComplianceState = {"document_path": document_path}
+    result = compliance_workflow.invoke(initial_state)
+    return result

backend/app/workflows/rag_workflow.py

@@ -0,0 +1,114 @@
+from typing import TypedDict, List
+from langgraph.graph import StateGraph, END
+
+
+class RagState(TypedDict):
+    query: str
+    query_embedding: List[float]
+    retrieved_docs: List[dict]
+    context: str
+    answer: str
+    sources: List[dict]
+
+
+def embed_query(state: RagState) -> dict:
+    """将查询转为向量"""
+    from app.services import embedding_service
+    embedding = embedding_service.embed_single(state["query"])
+    return {"query_embedding": embedding}
+
+
+def retrieve_docs(state: RagState) -> dict:
+    """向量检索"""
+    from app.services import milvus_service
+    from app.core.config import settings
+    docs = milvus_service.search(
+        state["query_embedding"],
+        top_k=settings.vector_top_k,
+    )
+    return {"retrieved_docs": docs[:settings.final_top_k]}
+
+
+def build_context(state: RagState) -> dict:
+    """构建上下文"""
+    context_parts = []
+    sources = []
+
+    for doc in state["retrieved_docs"]:
+        context_parts.append(f"【{doc['doc_name']} - {doc.get('clause_id', '')}】\n{doc['content']}")
+        sources.append({
+            "name": doc["doc_name"],
+            "clause": doc.get("clause_id"),
+        })
+
+    context = "\n\n".join(context_parts)
+    return {"context": context, "sources": sources}
+
+
+def generate_answer(state: RagState) -> dict:
+    """生成答案"""
+    from app.services import llm_service
+    prompt = f"""请根据以下法规内容回答用户问题，并在回答中标注引用的法规条款。
+
+法规内容：
+{state['context']}
+
+用户问题：{state['query']}
+
+请给出准确、简洁的回答，并引用相关法规条款。"""
+
+    answer = ""
+    for chunk in llm_service.generate_stream(prompt):
+        answer += chunk
+
+    return {"answer": answer}
+
+
+# 构建工作流图
+rag_graph = StateGraph(RagState)
+
+rag_graph.add_node("embed", embed_query)
+rag_graph.add_node("retrieve", retrieve_docs)
+rag_graph.add_node("build_context", build_context)
+rag_graph.add_node("generate", generate_answer)
+
+rag_graph.set_entry_point("embed")
+rag_graph.add_edge("embed", "retrieve")
+rag_graph.add_edge("retrieve", "build_context")
+rag_graph.add_edge("build_context", "generate")
+rag_graph.add_edge("generate", END)
+
+rag_workflow = rag_graph.compile()
+
+
+async def run_rag_workflow(query: str) -> RagState:
+    """运行RAG工作流"""
+    initial_state: RagState = {"query": query}
+    result = rag_workflow.invoke(initial_state)
+    return result
+
+
+def stream_rag_workflow(query: str):
+    """流式运行RAG工作流"""
+    from app.services import llm_service
+
+    # 先完成检索阶段
+    state: RagState = {"query": query}
+    state.update(embed_query(state))
+    state.update(retrieve_docs(state))
+    state.update(build_context(state))
+
+    # 流式生成阶段
+    prompt = f"""请根据以下法规内容回答用户问题，并在回答中标注引用的法规条款。
+
+法规内容：
+{state['context']}
+
+用户问题：{state['query']}
+
+请给出准确、简洁的回答，并引用相关法规条款。"""
+
+    for chunk in llm_service.generate_stream(prompt):
+        yield {"type": "chunk", "text": chunk}
+
+    yield {"type": "done", "sources": state["sources"]}

backend/data/raw/compliance_task-32e64724_test_doc.txt

@@ -0,0 +1 @@
+test content

backend/data/raw/doc-3b47abd7_requirement.txt

@@ -0,0 +1,2 @@
+apache-flink==1.13.2
+PyMySQL>=1.1.0

backend/data/raw/doc-9b01a78a_requirement.txt

@@ -0,0 +1,2 @@
+apache-flink==1.13.2
+PyMySQL>=1.1.0

backend/main.py

@@ -0,0 +1,19 @@
+"""Convenience launcher for the migrated backend app."""
+
+import uvicorn
+
+from app.config.settings import settings
+
+
+def main() -> None:
+    uvicorn.run(
+        "app.main:app",
+        host=settings.api_host,
+        port=settings.api_port,
+        reload=settings.debug,
+        log_level="info",
+    )
+
+
+if __name__ == "__main__":
+    main()

backend/pyproject.toml

@@ -0,0 +1,35 @@
+[project]
+name = "ai-regulations-backend"
+version = "0.1.0"
+description = "Migrated FastAPI backend for AI regulations demo"
+readme = "README.md"
+requires-python = ">=3.10"
+dependencies = [
+    "fastapi>=0.110.0",
+    "uvicorn[standard]>=0.27.0",
+    "python-multipart>=0.0.9",
+    "pydantic>=2.0.0",
+    "pydantic-settings>=2.0.0",
+    "python-dotenv>=1.0.0",
+    "loguru>=0.7.0",
+    "httpx>=0.25.0",
+    "tiktoken>=0.5.0",
+    "tenacity>=8.2.0",
+    "pymilvus>=2.4.0",
+    "minio>=7.1.0",
+    "pymupdf>=1.24.0",
+    "python-docx>=1.1.0",
+    "FlagEmbedding>=1.2.0",
+    "sentence-transformers>=2.2.0",
+    "torch>=2.0.0",
+    "numpy>=1.24.0",
+    "langchain>=0.1.0",
+    "langchain-milvus>=0.1.0",
+]
+
+[project.scripts]
+backend = "main:main"
+
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"

backend/requirements.txt

@@ -0,0 +1,29 @@
+fastapi>=0.110.0
+uvicorn[standard]>=0.27.0
+python-multipart>=0.0.9
+
+pydantic>=2.0.0
+pydantic-settings>=2.0.0
+python-dotenv>=1.0.0
+loguru>=0.7.0
+
+httpx>=0.25.0
+tiktoken>=0.5.0
+tenacity>=8.2.0
+
+pymilvus>=2.4.0
+minio>=7.1.0
+
+pymupdf>=1.24.0
+python-docx>=1.1.0
+
+FlagEmbedding>=1.2.0
+sentence-transformers>=2.2.0
+torch>=2.0.0
+numpy>=1.24.0
+
+langchain>=0.1.0
+langchain-milvus>=0.1.0
+
+pytest>=7.4.0
+pytest-asyncio>=0.21.0

backend/uv.lock

@@ -0,0 +1,8 @@
+version = 1
+revision = 3
+requires-python = ">=3.9"
+
+[[package]]
+name = "backend"
+version = "0.1.0"
+source = { virtual = "." }