feat: Add AI report generation with DIFY integration

- Add Users table for display name resolution from AD authentication
- Integrate DIFY AI service for report content generation
- Create docx assembly service with image embedding from MinIO
- Add REST API endpoints for report generation and download
- Add WebSocket notifications for generation progress
- Add frontend UI with progress modal and download functionality
- Add integration tests for report generation flow

Report sections (Traditional Chinese):
- 事件摘要 (Summary)
- 時間軸 (Timeline)
- 參與人員 (Participants)
- 處理過程 (Resolution Process)
- 目前狀態 (Current Status)
- 最終處置結果 (Final Resolution)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

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

app/modules/auth/models.py

@@ -2,6 +2,7 @@
 
 資料表結構：
 - user_sessions: 儲存使用者 session 資料，包含加密密碼用於自動刷新
+- users: 永久儲存使用者資訊 (用於報告生成時的姓名解析)
 """
 from sqlalchemy import Column, Integer, String, DateTime, Index
 from datetime import datetime
@@ -29,3 +30,39 @@ class UserSession(Base):
         DateTime, default=datetime.utcnow, nullable=False, comment="Last API request time"
     )
     created_at = Column(DateTime, default=datetime.utcnow, nullable=False)
+
+
+class User(Base):
+    """Permanent user information for display name resolution in reports
+
+    This table stores user information from AD API and persists even after
+    session expiration. Used for:
+    - Displaying user names (instead of emails) in generated reports
+    - Tracking user metadata (office location, job title)
+    """
+
+    __tablename__ = "users"
+
+    user_id = Column(
+        String(255), primary_key=True, comment="User email address (e.g., ymirliu@panjit.com.tw)"
+    )
+    display_name = Column(
+        String(255), nullable=False, comment="Display name from AD (e.g., 'ymirliu 劉念蓉')"
+    )
+    office_location = Column(
+        String(100), nullable=True, comment="Office location from AD (e.g., '高雄')"
+    )
+    job_title = Column(
+        String(100), nullable=True, comment="Job title from AD"
+    )
+    last_login_at = Column(
+        DateTime, nullable=True, comment="Last login timestamp"
+    )
+    created_at = Column(
+        DateTime, default=datetime.utcnow, nullable=False, comment="First login timestamp"
+    )
+
+    # Indexes
+    __table_args__ = (
+        Index("ix_users_display_name", "display_name"),
+    )

app/modules/auth/router.py

@@ -11,6 +11,7 @@ from app.modules.auth.schemas import LoginRequest, LoginResponse, LogoutResponse
 from app.modules.auth.services.ad_client import ad_auth_service
 from app.modules.auth.services.encryption import encryption_service
 from app.modules.auth.services.session_service import session_service
+from app.modules.auth.services.user_service import upsert_user
 from fastapi import Header
 from typing import Optional
 
@@ -30,10 +31,11 @@ async def login(request: LoginRequest, db: Session = Depends(get_db)):
 
     流程：
     1. 呼叫 AD API 驗證憑證
-    2. 加密密碼（用於自動刷新）
-    3. 生成 internal token (UUID)
-    4. 儲存 session 到資料庫
-    5. 回傳 internal token 和 display_name
+    2. 儲存/更新使用者資訊到 users 表（用於報告姓名解析）
+    3. 加密密碼（用於自動刷新）
+    4. 生成 internal token (UUID)
+    5. 儲存 session 到資料庫
+    6. 回傳 internal token 和 display_name
     """
     try:
         # Step 1: Authenticate with AD API
@@ -52,10 +54,19 @@ async def login(request: LoginRequest, db: Session = Depends(get_db)):
             detail="Authentication service unavailable",
         )
 
-    # Step 2: Encrypt password for future auto-refresh
+    # Step 2: Upsert user info for report generation (permanent storage)
+    upsert_user(
+        db=db,
+        user_id=ad_result["email"],
+        display_name=ad_result["username"],
+        office_location=ad_result.get("office_location"),
+        job_title=ad_result.get("job_title"),
+    )
+
+    # Step 3: Encrypt password for future auto-refresh
     encrypted_password = encryption_service.encrypt_password(request.password)
 
-    # Step 3 & 4: Generate internal token and create session
+    # Step 4 & 5: Generate internal token and create session
     user_session = session_service.create_session(
         db=db,
         username=request.username,

app/modules/auth/services/ad_client.py

@@ -31,6 +31,9 @@ class ADAuthService:
             Dict containing:
                 - token: AD authentication token
                 - username: Display name from AD
+                - email: User email address
+                - office_location: Office location (optional)
+                - job_title: Job title (optional)
                 - expires_at: Estimated token expiry datetime
 
         Raises:
@@ -58,6 +61,9 @@ class ADAuthService:
             ad_token = token_data.get("access_token")
             user_info = token_data.get("userInfo", {})
             display_name = user_info.get("name") or username
+            email = user_info.get("email") or username
+            office_location = user_info.get("officeLocation")
+            job_title = user_info.get("jobTitle")
 
             if not ad_token:
                 raise ValueError("No token received from AD API")
@@ -74,7 +80,14 @@ class ADAuthService:
                 # Fallback: assume 1 hour if not provided
                 expires_at = datetime.utcnow() + timedelta(hours=1)
 
-            return {"token": ad_token, "username": display_name, "expires_at": expires_at}
+            return {
+                "token": ad_token,
+                "username": display_name,
+                "email": email,
+                "office_location": office_location,
+                "job_title": job_title,
+                "expires_at": expires_at,
+            }
 
         except httpx.HTTPStatusError as e:
             # Authentication failed (401) or other HTTP errors

app/modules/auth/services/user_service.py

@@ -0,0 +1,89 @@
+"""User service for permanent user information storage
+
+This service handles upsert operations for the users table,
+which stores display names and metadata for report generation.
+"""
+from datetime import datetime
+from typing import Optional
+from sqlalchemy.orm import Session
+from app.modules.auth.models import User
+
+
+def upsert_user(
+    db: Session,
+    user_id: str,
+    display_name: str,
+    office_location: Optional[str] = None,
+    job_title: Optional[str] = None,
+) -> User:
+    """Create or update user record with AD information
+
+    This function is called on every successful login to keep
+    user information up to date. Uses SQLAlchemy merge for
+    atomic upsert operation.
+
+    Args:
+        db: Database session
+        user_id: User email address (primary key)
+        display_name: Display name from AD API
+        office_location: Office location from AD API (optional)
+        job_title: Job title from AD API (optional)
+
+    Returns:
+        User: The created or updated user record
+    """
+    # Check if user exists
+    existing_user = db.query(User).filter(User.user_id == user_id).first()
+
+    if existing_user:
+        # Update existing user
+        existing_user.display_name = display_name
+        existing_user.office_location = office_location
+        existing_user.job_title = job_title
+        existing_user.last_login_at = datetime.utcnow()
+        db.commit()
+        db.refresh(existing_user)
+        return existing_user
+    else:
+        # Create new user
+        new_user = User(
+            user_id=user_id,
+            display_name=display_name,
+            office_location=office_location,
+            job_title=job_title,
+            last_login_at=datetime.utcnow(),
+            created_at=datetime.utcnow(),
+        )
+        db.add(new_user)
+        db.commit()
+        db.refresh(new_user)
+        return new_user
+
+
+def get_user_by_id(db: Session, user_id: str) -> Optional[User]:
+    """Get user by user_id (email)
+
+    Args:
+        db: Database session
+        user_id: User email address
+
+    Returns:
+        User or None if not found
+    """
+    return db.query(User).filter(User.user_id == user_id).first()
+
+
+def get_display_name(db: Session, user_id: str) -> str:
+    """Get display name for a user, falling back to email if not found
+
+    Args:
+        db: Database session
+        user_id: User email address
+
+    Returns:
+        Display name or email address as fallback
+    """
+    user = get_user_by_id(db, user_id)
+    if user:
+        return user.display_name
+    return user_id  # Fallback to email if user not in database

app/modules/report_generation/__init__.py

@@ -0,0 +1,8 @@
+"""Report Generation Module
+
+AI-powered incident report generation using DIFY service.
+"""
+from app.modules.report_generation.models import GeneratedReport, ReportStatus
+from app.modules.report_generation.router import router
+
+__all__ = ["GeneratedReport", "ReportStatus", "router"]

app/modules/report_generation/dependencies.py

@@ -0,0 +1,53 @@
+"""Dependencies for report generation routes
+
+Provides permission checks for report-related operations.
+"""
+from fastapi import Depends, HTTPException, status
+from sqlalchemy.orm import Session
+
+from app.core.database import get_db
+from app.modules.auth import get_current_user
+from app.modules.chat_room.services.membership_service import membership_service
+from app.modules.chat_room.models import IncidentRoom
+
+
+def require_room_member(room_id: str, db: Session = Depends(get_db), current_user: dict = Depends(get_current_user)):
+    """Verify user is a member of the room
+
+    Args:
+        room_id: Room ID to check membership for
+        db: Database session
+        current_user: Current authenticated user
+
+    Returns:
+        dict with room_id and user_email
+
+    Raises:
+        HTTPException 404: If room not found
+        HTTPException 403: If user is not a member
+    """
+    user_email = current_user["username"]
+
+    # Check if room exists
+    room = db.query(IncidentRoom).filter(IncidentRoom.room_id == room_id).first()
+    if not room:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail="Room not found"
+        )
+
+    # Check if user is a member (or admin)
+    is_admin = membership_service.is_system_admin(user_email)
+    if not is_admin:
+        role = membership_service.get_user_role_in_room(db, room_id, user_email)
+        if not role:
+            raise HTTPException(
+                status_code=status.HTTP_403_FORBIDDEN,
+                detail="You are not a member of this room"
+            )
+
+    return {
+        "room_id": room_id,
+        "user_email": user_email,
+        "room": room,
+    }

app/modules/report_generation/models.py

@@ -0,0 +1,100 @@
+"""SQLAlchemy models for report generation
+
+Tables:
+- generated_reports: Stores report metadata and generation status
+"""
+from sqlalchemy import Column, String, Text, DateTime, Integer, ForeignKey, Index, JSON
+from sqlalchemy.orm import relationship
+from datetime import datetime
+import uuid
+import enum
+from app.core.database import Base
+
+
+class ReportStatus(str, enum.Enum):
+    """Report generation status"""
+    PENDING = "pending"
+    COLLECTING_DATA = "collecting_data"
+    GENERATING_CONTENT = "generating_content"
+    ASSEMBLING_DOCUMENT = "assembling_document"
+    COMPLETED = "completed"
+    FAILED = "failed"
+
+
+class GeneratedReport(Base):
+    """Generated report model for incident reports"""
+
+    __tablename__ = "generated_reports"
+
+    report_id = Column(
+        String(36), primary_key=True, default=lambda: str(uuid.uuid4()),
+        comment="Unique report identifier (UUID)"
+    )
+    room_id = Column(
+        String(36), ForeignKey("incident_rooms.room_id", ondelete="CASCADE"),
+        nullable=False, comment="Reference to incident room"
+    )
+
+    # Generation metadata
+    generated_by = Column(
+        String(255), nullable=False,
+        comment="User email who triggered report generation"
+    )
+    generated_at = Column(
+        DateTime, default=datetime.utcnow, nullable=False,
+        comment="Report generation timestamp"
+    )
+
+    # Status tracking
+    status = Column(
+        String(30), default=ReportStatus.PENDING.value, nullable=False,
+        comment="Current generation status"
+    )
+    error_message = Column(
+        Text, nullable=True,
+        comment="User-friendly error message if generation failed"
+    )
+
+    # DIFY AI metadata
+    dify_message_id = Column(
+        String(100), nullable=True,
+        comment="DIFY API message ID for tracking"
+    )
+    dify_conversation_id = Column(
+        String(100), nullable=True,
+        comment="DIFY conversation ID"
+    )
+    prompt_tokens = Column(
+        Integer, nullable=True,
+        comment="Number of prompt tokens used"
+    )
+    completion_tokens = Column(
+        Integer, nullable=True,
+        comment="Number of completion tokens used"
+    )
+
+    # Report content
+    report_title = Column(
+        String(255), nullable=True,
+        comment="Generated report title"
+    )
+    report_json = Column(
+        JSON, nullable=True,
+        comment="Parsed AI output as JSON"
+    )
+    docx_storage_path = Column(
+        String(500), nullable=True,
+        comment="Path to generated .docx file in MinIO or local storage"
+    )
+
+    # Relationship
+    room = relationship("IncidentRoom", backref="reports")
+
+    # Indexes
+    __table_args__ = (
+        Index("ix_generated_reports_room_date", "room_id", "generated_at"),
+        Index("ix_generated_reports_status", "status"),
+    )
+
+    def __repr__(self):
+        return f"<GeneratedReport {self.report_id} status={self.status}>"

app/modules/report_generation/prompts.py

@@ -0,0 +1,199 @@
+"""Prompt templates for DIFY AI report generation
+
+Contains the prompt construction logic for building the user query
+sent to DIFY Chat API.
+"""
+from typing import List, Dict, Any
+from datetime import datetime
+
+
+INCIDENT_TYPE_MAP = {
+    "equipment_failure": "設備故障",
+    "material_shortage": "物料短缺",
+    "quality_issue": "品質問題",
+    "other": "其他",
+}
+
+SEVERITY_MAP = {
+    "low": "低",
+    "medium": "中",
+    "high": "高",
+    "critical": "緊急",
+}
+
+STATUS_MAP = {
+    "active": "處理中",
+    "resolved": "已解決",
+    "archived": "已封存",
+}
+
+MEMBER_ROLE_MAP = {
+    "owner": "建立者",
+    "editor": "編輯者",
+    "viewer": "檢視者",
+}
+
+
+def build_report_prompt(
+    room_data: Dict[str, Any],
+    messages: List[Dict[str, Any]],
+    members: List[Dict[str, Any]],
+    files: List[Dict[str, Any]],
+) -> str:
+    """Build the complete prompt for DIFY report generation
+
+    Args:
+        room_data: Room metadata (title, type, severity, status, etc.)
+        messages: List of messages with sender_name, content, created_at
+        members: List of members with display_name, role
+        files: List of files with filename, file_type, uploaded_at, uploader_name
+
+    Returns:
+        Formatted prompt string for DIFY API
+    """
+    sections = []
+
+    # Section 1: Event Information
+    sections.append(_format_room_info(room_data))
+
+    # Section 2: Participants
+    sections.append(_format_members(members))
+
+    # Section 3: Message Timeline
+    sections.append(_format_messages(messages))
+
+    # Section 4: File Attachments
+    sections.append(_format_files(files))
+
+    # Section 5: Instructions
+    sections.append(_format_instructions())
+
+    return "\n\n".join(sections)
+
+
+def _format_room_info(room_data: Dict[str, Any]) -> str:
+    """Format room metadata section"""
+    incident_type = INCIDENT_TYPE_MAP.get(
+        room_data.get("incident_type"), room_data.get("incident_type")
+    )
+    severity = SEVERITY_MAP.get(room_data.get("severity"), room_data.get("severity"))
+    status = STATUS_MAP.get(room_data.get("status"), room_data.get("status"))
+
+    created_at = room_data.get("created_at")
+    if isinstance(created_at, datetime):
+        created_at = created_at.strftime("%Y-%m-%d %H:%M")
+
+    resolved_at = room_data.get("resolved_at")
+    if isinstance(resolved_at, datetime):
+        resolved_at = resolved_at.strftime("%Y-%m-%d %H:%M")
+    elif resolved_at is None:
+        resolved_at = "尚未解決"
+
+    lines = [
+        "## 事件資訊",
+        f"- 標題: {room_data.get('title', '未命名')}",
+        f"- 類型: {incident_type}",
+        f"- 嚴重程度: {severity}",
+        f"- 目前狀態: {status}",
+        f"- 發生地點: {room_data.get('location', '未指定')}",
+        f"- 建立時間: {created_at}",
+        f"- 解決時間: {resolved_at}",
+    ]
+
+    if room_data.get("description"):
+        lines.append(f"- 事件描述: {room_data['description']}")
+
+    if room_data.get("resolution_notes"):
+        lines.append(f"- 解決備註: {room_data['resolution_notes']}")
+
+    return "\n".join(lines)
+
+
+def _format_members(members: List[Dict[str, Any]]) -> str:
+    """Format participants section"""
+    lines = ["## 參與人員"]
+
+    if not members:
+        lines.append("無參與人員記錄")
+        return "\n".join(lines)
+
+    for member in members:
+        display_name = member.get("display_name") or member.get("user_id", "未知")
+        role = MEMBER_ROLE_MAP.get(member.get("role"), member.get("role", "成員"))
+        lines.append(f"- {display_name} ({role})")
+
+    return "\n".join(lines)
+
+
+def _format_messages(messages: List[Dict[str, Any]]) -> str:
+    """Format message timeline section"""
+    lines = ["## 對話記錄"]
+
+    if not messages:
+        lines.append("無對話記錄")
+        return "\n".join(lines)
+
+    for msg in messages:
+        sender = msg.get("sender_name") or msg.get("sender_id", "未知")
+        content = msg.get("content", "")
+        msg_type = msg.get("message_type", "text")
+
+        created_at = msg.get("created_at")
+        if isinstance(created_at, datetime):
+            time_str = created_at.strftime("%Y-%m-%d %H:%M")
+        else:
+            time_str = str(created_at) if created_at else "未知時間"
+
+        # Handle different message types
+        if msg_type == "file":
+            file_name = msg.get("file_name", "附件")
+            lines.append(f"[{time_str}] {sender}: [上傳檔案: {file_name}]")
+        elif msg_type == "image":
+            lines.append(f"[{time_str}] {sender}: [上傳圖片]")
+        elif msg_type == "system":
+            lines.append(f"[{time_str}] [系統]: {content}")
+        else:
+            lines.append(f"[{time_str}] {sender}: {content}")
+
+    return "\n".join(lines)
+
+
+def _format_files(files: List[Dict[str, Any]]) -> str:
+    """Format file attachments section"""
+    lines = ["## 附件清單"]
+
+    if not files:
+        lines.append("無附件")
+        return "\n".join(lines)
+
+    for f in files:
+        filename = f.get("filename", "未命名檔案")
+        file_type = f.get("file_type", "file")
+        uploader = f.get("uploader_name") or f.get("uploaded_by", "未知")
+
+        uploaded_at = f.get("uploaded_at")
+        if isinstance(uploaded_at, datetime):
+            time_str = uploaded_at.strftime("%Y-%m-%d %H:%M")
+        else:
+            time_str = str(uploaded_at) if uploaded_at else ""
+
+        type_label = "圖片" if file_type == "image" else "檔案"
+        lines.append(f"- [{type_label}] {filename} (由 {uploader} 於 {time_str} 上傳)")
+
+    return "\n".join(lines)
+
+
+def _format_instructions() -> str:
+    """Format generation instructions"""
+    return """## 報告生成指示
+
+請根據以上資料，生成一份結構化的事件報告。報告必須為 JSON 格式，包含以下欄位：
+
+1. **summary**: 事件摘要 (50-100字)
+2. **timeline**: 按時間順序的事件時間軸
+3. **participants**: 參與人員及其角色
+4. **resolution_process**: 詳細的處理過程描述
+5. **current_status**: 目前狀態說明
+6. **final_resolution**: 最終處置結果（若已解決）
+
+請直接輸出 JSON，不要包含其他說明文字。"""

app/modules/report_generation/router.py

@@ -0,0 +1,445 @@
+"""API routes for report generation
+
+FastAPI router with all report-related endpoints:
+- POST /api/rooms/{room_id}/reports/generate - Trigger report generation
+- GET /api/rooms/{room_id}/reports - List reports for a room
+- GET /api/rooms/{room_id}/reports/{report_id} - Get report status/metadata
+- GET /api/rooms/{room_id}/reports/{report_id}/download - Download report .docx
+"""
+import logging
+from fastapi import APIRouter, Depends, HTTPException, status, BackgroundTasks
+from fastapi.responses import StreamingResponse
+from sqlalchemy.orm import Session
+from typing import Optional
+import io
+
+from app.core.database import get_db
+from app.core.config import get_settings
+from app.core.minio_client import get_minio_client
+from app.modules.auth import get_current_user
+from app.modules.report_generation import schemas
+from app.modules.report_generation.models import GeneratedReport, ReportStatus
+from app.modules.report_generation.dependencies import require_room_member
+from app.modules.report_generation.services.report_data_service import ReportDataService
+from app.modules.report_generation.services.dify_client import (
+    dify_service,
+    DifyAPIError,
+    DifyJSONParseError,
+    DifyValidationError,
+)
+from app.modules.report_generation.services.docx_service import docx_service
+from app.modules.report_generation.prompts import build_report_prompt
+from app.modules.realtime.websocket_manager import manager as ws_manager
+
+settings = get_settings()
+logger = logging.getLogger(__name__)
+
+
+async def _broadcast_report_progress(
+    room_id: str,
+    report_id: str,
+    status: str,
+    message: str,
+    error: str = None,
+):
+    """Broadcast report generation progress via WebSocket
+
+    Args:
+        room_id: Room ID to broadcast to
+        report_id: Report ID
+        status: Current status (pending, collecting_data, generating_content, etc.)
+        message: Human-readable progress message
+        error: Error message if failed
+    """
+    payload = {
+        "type": "report_progress",
+        "report_id": report_id,
+        "room_id": room_id,
+        "status": status,
+        "message": message,
+    }
+    if error:
+        payload["error"] = error
+
+    await ws_manager.broadcast_to_room(room_id, payload)
+
+router = APIRouter(prefix="/api/rooms/{room_id}/reports", tags=["Report Generation"])
+
+
+@router.post("/generate", response_model=schemas.ReportGenerateResponse, status_code=status.HTTP_202_ACCEPTED)
+async def generate_report(
+    room_id: str,
+    request: schemas.ReportGenerateRequest = None,
+    background_tasks: BackgroundTasks = None,
+    db: Session = Depends(get_db),
+    room_context: dict = Depends(require_room_member),
+):
+    """Trigger AI report generation for a room
+
+    This endpoint starts an async report generation process:
+    1. Collects room data (messages, members, files)
+    2. Sends data to DIFY AI for content generation
+    3. Assembles .docx document with AI content and images
+    4. Stores report in MinIO
+
+    Args:
+        room_id: Room ID to generate report for
+        request: Optional generation parameters
+
+    Returns:
+        Report ID and initial status
+    """
+    if request is None:
+        request = schemas.ReportGenerateRequest()
+
+    user_email = room_context["user_email"]
+
+    # Create report record with pending status
+    report = GeneratedReport(
+        room_id=room_id,
+        generated_by=user_email,
+        status=ReportStatus.PENDING.value,
+    )
+    db.add(report)
+    db.commit()
+    db.refresh(report)
+
+    # Start background generation task
+    background_tasks.add_task(
+        _generate_report_task,
+        report_id=report.report_id,
+        room_id=room_id,
+        include_images=request.include_images,
+        include_file_list=request.include_file_list,
+    )
+
+    return schemas.ReportGenerateResponse(
+        report_id=report.report_id,
+        status=schemas.ReportStatus.PENDING,
+        message="Report generation started",
+    )
+
+
+@router.get("", response_model=schemas.ReportListResponse)
+async def list_reports(
+    room_id: str,
+    db: Session = Depends(get_db),
+    room_context: dict = Depends(require_room_member),
+):
+    """List all reports for a room
+
+    Args:
+        room_id: Room ID to list reports for
+
+    Returns:
+        List of report summaries sorted by generation time (newest first)
+    """
+    reports = (
+        db.query(GeneratedReport)
+        .filter(GeneratedReport.room_id == room_id)
+        .order_by(GeneratedReport.generated_at.desc())
+        .all()
+    )
+
+    items = [
+        schemas.ReportListItem(
+            report_id=r.report_id,
+            generated_at=r.generated_at,
+            generated_by=r.generated_by,
+            status=schemas.ReportStatus(r.status),
+            report_title=r.report_title,
+        )
+        for r in reports
+    ]
+
+    return schemas.ReportListResponse(
+        reports=items,
+        total=len(items),
+    )
+
+
+@router.get("/{report_id}", response_model=schemas.ReportStatusResponse)
+async def get_report_status(
+    room_id: str,
+    report_id: str,
+    db: Session = Depends(get_db),
+    room_context: dict = Depends(require_room_member),
+):
+    """Get report status and metadata
+
+    Args:
+        room_id: Room ID
+        report_id: Report ID to get status for
+
+    Returns:
+        Report metadata including status, token usage, etc.
+    """
+    report = (
+        db.query(GeneratedReport)
+        .filter(
+            GeneratedReport.report_id == report_id,
+            GeneratedReport.room_id == room_id,
+        )
+        .first()
+    )
+
+    if not report:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail="Report not found",
+        )
+
+    return schemas.ReportStatusResponse(
+        report_id=report.report_id,
+        room_id=report.room_id,
+        generated_by=report.generated_by,
+        generated_at=report.generated_at,
+        status=schemas.ReportStatus(report.status),
+        error_message=report.error_message,
+        report_title=report.report_title,
+        prompt_tokens=report.prompt_tokens,
+        completion_tokens=report.completion_tokens,
+    )
+
+
+@router.get("/{report_id}/download")
+async def download_report(
+    room_id: str,
+    report_id: str,
+    db: Session = Depends(get_db),
+    room_context: dict = Depends(require_room_member),
+):
+    """Download generated report as .docx
+
+    Args:
+        room_id: Room ID
+        report_id: Report ID to download
+
+    Returns:
+        StreamingResponse with .docx file
+    """
+    report = (
+        db.query(GeneratedReport)
+        .filter(
+            GeneratedReport.report_id == report_id,
+            GeneratedReport.room_id == room_id,
+        )
+        .first()
+    )
+
+    if not report:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail="Report not found",
+        )
+
+    if report.status != ReportStatus.COMPLETED.value:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=f"Report is not ready for download. Status: {report.status}",
+        )
+
+    if not report.docx_storage_path:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail="Report file not found in storage",
+        )
+
+    # Download from MinIO
+    try:
+        minio_client = get_minio_client()
+        response = minio_client.get_object(
+            settings.MINIO_BUCKET,
+            report.docx_storage_path,
+        )
+
+        # Read file content
+        content = response.read()
+        response.close()
+        response.release_conn()
+
+        # Create filename from report title or ID
+        filename = report.report_title or f"report_{report.report_id[:8]}"
+        filename = f"{filename}.docx"
+
+        return StreamingResponse(
+            io.BytesIO(content),
+            media_type="application/vnd.openxmlformats-officedocument.wordprocessingml.document",
+            headers={
+                "Content-Disposition": f'attachment; filename="{filename}"',
+            },
+        )
+
+    except Exception as e:
+        logger.error(f"Failed to download report from MinIO: {e}")
+        raise HTTPException(
+            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+            detail="Failed to retrieve report file",
+        )
+
+
+async def _generate_report_task(
+    report_id: str,
+    room_id: str,
+    include_images: bool = True,
+    include_file_list: bool = True,
+):
+    """Background task to generate report
+
+    This task:
+    1. Updates status to COLLECTING_DATA + WebSocket notification
+    2. Collects room data
+    3. Updates status to GENERATING_CONTENT + WebSocket notification
+    4. Calls DIFY AI for content
+    5. Updates status to ASSEMBLING_DOCUMENT + WebSocket notification
+    6. Creates .docx document
+    7. Uploads to MinIO
+    8. Updates status to COMPLETED (or FAILED) + WebSocket notification
+    """
+    from app.core.database import SessionLocal
+
+    db = SessionLocal()
+
+    try:
+        report = db.query(GeneratedReport).filter(
+            GeneratedReport.report_id == report_id
+        ).first()
+
+        if not report:
+            logger.error(f"Report not found: {report_id}")
+            return
+
+        # Phase 1: Collecting Data
+        report.status = ReportStatus.COLLECTING_DATA.value
+        db.commit()
+        await _broadcast_report_progress(
+            room_id, report_id, "collecting_data", "正在收集聊天室資料..."
+        )
+
+        data_service = ReportDataService(db)
+        room_data = data_service.collect_room_data(room_id)
+
+        if not room_data:
+            report.status = ReportStatus.FAILED.value
+            report.error_message = "Room not found"
+            db.commit()
+            await _broadcast_report_progress(
+                room_id, report_id, "failed", "報告生成失敗", error="找不到聊天室"
+            )
+            return
+
+        # Convert to dict for prompt builder
+        prompt_data = data_service.to_prompt_dict(room_data)
+
+        # Phase 2: Generating Content
+        report.status = ReportStatus.GENERATING_CONTENT.value
+        db.commit()
+        await _broadcast_report_progress(
+            room_id, report_id, "generating_content", "AI 正在分析並生成報告內容..."
+        )
+
+        # Build prompt and call DIFY
+        prompt = build_report_prompt(
+            room_data=prompt_data["room_data"],
+            messages=prompt_data["messages"],
+            members=prompt_data["members"],
+            files=prompt_data["files"],
+        )
+
+        try:
+            dify_response = await dify_service.generate_report(prompt, room_id)
+        except (DifyAPIError, DifyJSONParseError, DifyValidationError) as e:
+            report.status = ReportStatus.FAILED.value
+            report.error_message = str(e)
+            db.commit()
+            await _broadcast_report_progress(
+                room_id, report_id, "failed", "AI 生成失敗", error=str(e)
+            )
+            logger.error(f"DIFY error for report {report_id}: {e}")
+            return
+
+        # Save AI response data
+        report.dify_message_id = dify_response.message_id
+        report.dify_conversation_id = dify_response.conversation_id
+        report.prompt_tokens = dify_response.prompt_tokens
+        report.completion_tokens = dify_response.completion_tokens
+        report.report_json = dify_response.parsed_json
+
+        # Extract title from summary
+        ai_content = dify_response.parsed_json
+        if ai_content and "summary" in ai_content:
+            summary_content = ai_content["summary"].get("content", "")
+            # Use first 50 chars of summary as title
+            report.report_title = summary_content[:50] + "..." if len(summary_content) > 50 else summary_content
+
+        db.commit()
+
+        # Phase 3: Assembling Document
+        report.status = ReportStatus.ASSEMBLING_DOCUMENT.value
+        db.commit()
+        await _broadcast_report_progress(
+            room_id, report_id, "assembling_document", "正在組裝報告文件..."
+        )
+
+        try:
+            docx_data = docx_service.create_report(
+                room_data=prompt_data["room_data"],
+                ai_content=ai_content,
+                files=prompt_data["files"],
+                include_images=include_images,
+                include_file_list=include_file_list,
+            )
+        except Exception as e:
+            report.status = ReportStatus.FAILED.value
+            report.error_message = f"Document assembly failed: {str(e)}"
+            db.commit()
+            await _broadcast_report_progress(
+                room_id, report_id, "failed", "文件組裝失敗", error=str(e)
+            )
+            logger.error(f"Document assembly error for report {report_id}: {e}")
+            return
+
+        # Upload to MinIO
+        storage_path = docx_service.upload_report(
+            report_data=docx_data,
+            room_id=room_id,
+            report_id=report_id,
+        )
+
+        if not storage_path:
+            report.status = ReportStatus.FAILED.value
+            report.error_message = "Failed to upload report to storage"
+            db.commit()
+            await _broadcast_report_progress(
+                room_id, report_id, "failed", "報告上傳失敗", error="無法上傳到儲存空間"
+            )
+            return
+
+        # Phase 4: Completed
+        report.docx_storage_path = storage_path
+        report.status = ReportStatus.COMPLETED.value
+        db.commit()
+        await _broadcast_report_progress(
+            room_id, report_id, "completed", "報告生成完成！"
+        )
+
+        logger.info(f"Report generation completed: {report_id}")
+
+    except Exception as e:
+        logger.error(f"Unexpected error generating report {report_id}: {e}")
+        try:
+            report = db.query(GeneratedReport).filter(
+                GeneratedReport.report_id == report_id
+            ).first()
+            if report:
+                report.status = ReportStatus.FAILED.value
+                report.error_message = f"Unexpected error: {str(e)}"
+                db.commit()
+            await _broadcast_report_progress(
+                room_id, report_id, "failed", "發生未預期的錯誤", error=str(e)
+            )
+        except Exception:
+            pass
+
+    finally:
+        db.close()

app/modules/report_generation/schemas.py

@@ -0,0 +1,105 @@
+"""Pydantic schemas for report generation API
+
+Request and response models for the report generation endpoints.
+"""
+from pydantic import BaseModel, Field
+from typing import Optional, List
+from datetime import datetime
+from enum import Enum
+
+
+class ReportStatus(str, Enum):
+    """Report generation status"""
+    PENDING = "pending"
+    COLLECTING_DATA = "collecting_data"
+    GENERATING_CONTENT = "generating_content"
+    ASSEMBLING_DOCUMENT = "assembling_document"
+    COMPLETED = "completed"
+    FAILED = "failed"
+
+
+# Request Schemas
+class ReportGenerateRequest(BaseModel):
+    """Request to generate a report (optional parameters)"""
+    include_images: bool = Field(default=True, description="Whether to embed images in the report")
+    include_file_list: bool = Field(default=True, description="Whether to include file attachment list")
+
+
+# Response Schemas
+class ReportGenerateResponse(BaseModel):
+    """Response after triggering report generation"""
+    report_id: str = Field(..., description="Unique report identifier")
+    status: ReportStatus = Field(..., description="Initial status (typically 'pending')")
+    message: str = Field(default="Report generation started", description="Status message")
+
+
+class ReportStatusResponse(BaseModel):
+    """Full report metadata response"""
+    report_id: str
+    room_id: str
+    generated_by: str
+    generated_at: datetime
+    status: ReportStatus
+    error_message: Optional[str] = None
+    report_title: Optional[str] = None
+    prompt_tokens: Optional[int] = None
+    completion_tokens: Optional[int] = None
+
+    class Config:
+        from_attributes = True
+
+
+class ReportListItem(BaseModel):
+    """Report item in list response"""
+    report_id: str
+    generated_at: datetime
+    generated_by: str
+    status: ReportStatus
+    report_title: Optional[str] = None
+
+    class Config:
+        from_attributes = True
+
+
+class ReportListResponse(BaseModel):
+    """List of reports for a room"""
+    reports: List[ReportListItem]
+    total: int
+
+
+# AI Report Content Schemas (validated JSON from DIFY)
+class TimelineEvent(BaseModel):
+    """Single event in timeline"""
+    time: str = Field(..., description="Time of event (HH:MM or YYYY-MM-DD HH:MM)")
+    description: str = Field(..., description="Event description")
+
+
+class ParticipantInfo(BaseModel):
+    """Participant information"""
+    name: str = Field(..., description="Participant name")
+    role: str = Field(..., description="Role in incident (e.g., 事件發起人, 維修負責人)")
+
+
+class AIReportContent(BaseModel):
+    """Validated JSON schema from DIFY AI response"""
+    summary: dict = Field(..., description="Event summary with 'content' field")
+    timeline: dict = Field(..., description="Timeline with 'events' list")
+    participants: dict = Field(..., description="Participants with 'members' list")
+    resolution_process: dict = Field(..., description="Resolution process with 'content' field")
+    current_status: dict = Field(..., description="Current status with 'status' and 'description' fields")
+    final_resolution: dict = Field(..., description="Final resolution with 'has_resolution' and 'content' fields")
+
+    @classmethod
+    def validate_structure(cls, data: dict) -> bool:
+        """Validate the basic structure of AI response"""
+        required_keys = ["summary", "timeline", "participants", "resolution_process", "current_status", "final_resolution"]
+        for key in required_keys:
+            if key not in data:
+                return False
+        return True
+
+
+# Error Response
+class ErrorResponse(BaseModel):
+    """Error response"""
+    detail: str

app/modules/report_generation/services/__init__.py

@@ -0,0 +1,13 @@
+"""Report generation services"""
+from app.modules.report_generation.services.dify_client import dify_service, DifyService
+from app.modules.report_generation.services.report_data_service import ReportDataService, RoomReportData
+from app.modules.report_generation.services.docx_service import docx_service, DocxAssemblyService
+
+__all__ = [
+    "dify_service",
+    "DifyService",
+    "ReportDataService",
+    "RoomReportData",
+    "docx_service",
+    "DocxAssemblyService",
+]

app/modules/report_generation/services/dify_client.py

@@ -0,0 +1,261 @@
+"""DIFY AI Service Client
+
+Handles communication with DIFY Chat API for report generation.
+"""
+import json
+import re
+import httpx
+from typing import Dict, Any, Optional
+from dataclasses import dataclass
+from app.core.config import get_settings
+
+settings = get_settings()
+
+
+@dataclass
+class DifyResponse:
+    """DIFY API response container"""
+    message_id: str
+    conversation_id: str
+    answer: str
+    parsed_json: Optional[Dict[str, Any]] = None
+    prompt_tokens: Optional[int] = None
+    completion_tokens: Optional[int] = None
+
+
+class DifyServiceError(Exception):
+    """Base exception for DIFY service errors"""
+    pass
+
+
+class DifyAPIError(DifyServiceError):
+    """API request failed"""
+    pass
+
+
+class DifyJSONParseError(DifyServiceError):
+    """Failed to parse JSON from AI response"""
+    pass
+
+
+class DifyValidationError(DifyServiceError):
+    """Response JSON doesn't match expected schema"""
+    pass
+
+
+class DifyService:
+    """DIFY Chat API client for report generation"""
+
+    REQUIRED_SECTIONS = [
+        "summary",
+        "timeline",
+        "participants",
+        "resolution_process",
+        "current_status",
+        "final_resolution",
+    ]
+
+    def __init__(self):
+        self.base_url = settings.DIFY_BASE_URL.rstrip("/")
+        self.api_key = settings.DIFY_API_KEY
+        self.timeout = settings.DIFY_TIMEOUT_SECONDS
+        self._client = httpx.AsyncClient(timeout=self.timeout)
+
+    async def generate_report(self, prompt: str, room_id: str) -> DifyResponse:
+        """Send prompt to DIFY and get AI-generated report content
+
+        Args:
+            prompt: Constructed prompt with room data and instructions
+            room_id: Room ID used as user identifier for tracking
+
+        Returns:
+            DifyResponse with parsed JSON content
+
+        Raises:
+            DifyAPIError: If API request fails
+            DifyJSONParseError: If response is not valid JSON
+            DifyValidationError: If JSON doesn't contain required sections
+        """
+        url = f"{self.base_url}/chat-messages"
+        headers = {
+            "Authorization": f"Bearer {self.api_key}",
+            "Content-Type": "application/json",
+        }
+        payload = {
+            "inputs": {},
+            "query": prompt,
+            "response_mode": "blocking",
+            "conversation_id": "",  # New conversation each time
+            "user": room_id,  # Use room_id for tracking
+        }
+
+        try:
+            response = await self._client.post(url, headers=headers, json=payload)
+            response.raise_for_status()
+        except httpx.TimeoutException as e:
+            raise DifyAPIError(f"DIFY API timeout after {self.timeout}s") from e
+        except httpx.HTTPStatusError as e:
+            error_detail = ""
+            try:
+                error_detail = e.response.text[:500]
+            except Exception:
+                pass
+            raise DifyAPIError(
+                f"DIFY API error: {e.response.status_code} - {error_detail}"
+            ) from e
+        except httpx.RequestError as e:
+            raise DifyAPIError(f"DIFY API connection error: {str(e)}") from e
+
+        # Parse API response
+        try:
+            data = response.json()
+        except json.JSONDecodeError as e:
+            raise DifyAPIError(f"Invalid JSON response from DIFY: {str(e)}") from e
+
+        # Extract fields from response
+        message_id = data.get("message_id", "")
+        conversation_id = data.get("conversation_id", "")
+        answer = data.get("answer", "")
+
+        # Extract token usage from metadata
+        metadata = data.get("metadata", {})
+        usage = metadata.get("usage", {})
+        prompt_tokens = usage.get("prompt_tokens")
+        completion_tokens = usage.get("completion_tokens")
+
+        if not answer:
+            raise DifyAPIError("Empty answer from DIFY API")
+
+        # Parse JSON from answer
+        parsed_json = self._extract_json(answer)
+
+        # Validate required sections
+        self._validate_schema(parsed_json)
+
+        return DifyResponse(
+            message_id=message_id,
+            conversation_id=conversation_id,
+            answer=answer,
+            parsed_json=parsed_json,
+            prompt_tokens=prompt_tokens,
+            completion_tokens=completion_tokens,
+        )
+
+    def _extract_json(self, text: str) -> Dict[str, Any]:
+        """Extract JSON from AI response text
+
+        Handles cases where:
+        1. Response is pure JSON
+        2. JSON is wrapped in markdown code blocks
+        3. JSON is embedded in other text
+
+        Args:
+            text: Raw text from AI response
+
+        Returns:
+            Parsed JSON as dictionary
+
+        Raises:
+            DifyJSONParseError: If no valid JSON found
+        """
+        text = text.strip()
+
+        # Try 1: Direct parse (pure JSON)
+        try:
+            return json.loads(text)
+        except json.JSONDecodeError:
+            pass
+
+        # Try 2: Extract from markdown code blocks ```json ... ``` or ``` ... ```
+        code_block_pattern = r"```(?:json)?\s*([\s\S]*?)\s*```"
+        matches = re.findall(code_block_pattern, text)
+        for match in matches:
+            try:
+                return json.loads(match.strip())
+            except json.JSONDecodeError:
+                continue
+
+        # Try 3: Find JSON object in text (first { to last })
+        json_start = text.find("{")
+        json_end = text.rfind("}")
+        if json_start != -1 and json_end != -1 and json_end > json_start:
+            try:
+                return json.loads(text[json_start : json_end + 1])
+            except json.JSONDecodeError:
+                pass
+
+        raise DifyJSONParseError(
+            f"Could not extract valid JSON from AI response. Response preview: {text[:200]}..."
+        )
+
+    def _validate_schema(self, data: Dict[str, Any]) -> None:
+        """Validate that response contains all required sections
+
+        Args:
+            data: Parsed JSON dictionary
+
+        Raises:
+            DifyValidationError: If required sections are missing
+        """
+        missing = [key for key in self.REQUIRED_SECTIONS if key not in data]
+        if missing:
+            raise DifyValidationError(
+                f"AI response missing required sections: {', '.join(missing)}"
+            )
+
+        # Validate summary has content
+        summary = data.get("summary", {})
+        if not isinstance(summary, dict) or not summary.get("content"):
+            raise DifyValidationError("summary section missing 'content' field")
+
+        # Validate timeline has events list
+        timeline = data.get("timeline", {})
+        if not isinstance(timeline, dict) or not isinstance(
+            timeline.get("events"), list
+        ):
+            raise DifyValidationError("timeline section missing 'events' list")
+
+        # Validate participants has members list
+        participants = data.get("participants", {})
+        if not isinstance(participants, dict) or not isinstance(
+            participants.get("members"), list
+        ):
+            raise DifyValidationError("participants section missing 'members' list")
+
+        # Validate resolution_process has content
+        resolution = data.get("resolution_process", {})
+        if not isinstance(resolution, dict) or "content" not in resolution:
+            raise DifyValidationError(
+                "resolution_process section missing 'content' field"
+            )
+
+        # Validate current_status has status and description
+        current_status = data.get("current_status", {})
+        if not isinstance(current_status, dict):
+            raise DifyValidationError("current_status must be a dictionary")
+        if "status" not in current_status or "description" not in current_status:
+            raise DifyValidationError(
+                "current_status section missing 'status' or 'description' field"
+            )
+
+        # Validate final_resolution has has_resolution (content required only if has_resolution=true)
+        final = data.get("final_resolution", {})
+        if not isinstance(final, dict):
+            raise DifyValidationError("final_resolution must be a dictionary")
+        if "has_resolution" not in final:
+            raise DifyValidationError(
+                "final_resolution section missing 'has_resolution' field"
+            )
+        # content is required only when has_resolution is true
+        if final.get("has_resolution") and "content" not in final:
+            raise DifyValidationError(
+                "final_resolution section missing 'content' field when has_resolution is true"
+            )
+
+    async def close(self):
+        """Close HTTP client"""
+        await self._client.aclose()
+
+
+# Singleton instance
+dify_service = DifyService()

app/modules/report_generation/services/docx_service.py

@@ -0,0 +1,445 @@
+"""Document Assembly Service
+
+Creates .docx reports using python-docx with:
+- Title and metadata header
+- AI-generated content sections
+- Embedded images from MinIO
+- File attachment list
+"""
+import io
+import logging
+from typing import Dict, Any, List, Optional
+from datetime import datetime
+from docx import Document
+from docx.shared import Inches, Pt, RGBColor
+from docx.enum.text import WD_ALIGN_PARAGRAPH
+from docx.enum.style import WD_STYLE_TYPE
+
+from app.core.config import get_settings
+from app.core.minio_client import get_minio_client
+from app.modules.file_storage.services.minio_service import upload_file
+
+settings = get_settings()
+logger = logging.getLogger(__name__)
+
+
+# Constants for document styling
+TITLE_SIZE = Pt(18)
+HEADING_SIZE = Pt(14)
+BODY_SIZE = Pt(11)
+CHINESE_FONT = "Microsoft JhengHei"
+FALLBACK_FONT = "Arial"
+
+
+class DocxAssemblyService:
+    """Service to create and assemble .docx incident reports"""
+
+    # Mapping of status values to Chinese labels
+    STATUS_MAP = {
+        "active": "處理中",
+        "resolved": "已解決",
+        "archived": "已封存",
+    }
+
+    INCIDENT_TYPE_MAP = {
+        "equipment_failure": "設備故障",
+        "material_shortage": "物料短缺",
+        "quality_issue": "品質問題",
+        "other": "其他",
+    }
+
+    SEVERITY_MAP = {
+        "low": "低",
+        "medium": "中",
+        "high": "高",
+        "critical": "緊急",
+    }
+
+    def __init__(self):
+        self.minio_client = get_minio_client()
+        self.bucket = settings.MINIO_BUCKET
+
+    def create_report(
+        self,
+        room_data: Dict[str, Any],
+        ai_content: Dict[str, Any],
+        files: List[Dict[str, Any]],
+        include_images: bool = True,
+        include_file_list: bool = True,
+    ) -> io.BytesIO:
+        """Create a complete incident report document
+
+        Args:
+            room_data: Room metadata (title, type, severity, status, etc.)
+            ai_content: AI-generated content (summary, timeline, participants, etc.)
+            files: List of files with metadata
+            include_images: Whether to embed images in the report
+            include_file_list: Whether to include file attachment list
+
+        Returns:
+            BytesIO object containing the .docx file
+        """
+        doc = Document()
+
+        # Configure document styles
+        self._setup_styles(doc)
+
+        # Add title
+        self._add_title(doc, room_data)
+
+        # Add metadata table
+        self._add_metadata_table(doc, room_data)
+
+        # Add AI-generated sections
+        self._add_summary_section(doc, ai_content)
+        self._add_timeline_section(doc, ai_content)
+        self._add_participants_section(doc, ai_content)
+        self._add_resolution_process_section(doc, ai_content)
+        self._add_current_status_section(doc, ai_content)
+        self._add_final_resolution_section(doc, ai_content)
+
+        # Add images if requested
+        if include_images and files:
+            self._add_images_section(doc, files)
+
+        # Add file attachment list if requested
+        if include_file_list and files:
+            self._add_file_list_section(doc, files)
+
+        # Save to BytesIO
+        output = io.BytesIO()
+        doc.save(output)
+        output.seek(0)
+
+        return output
+
+    def _setup_styles(self, doc: Document):
+        """Configure document styles"""
+        # Configure default paragraph style
+        style = doc.styles["Normal"]
+        font = style.font
+        font.name = CHINESE_FONT
+        font.size = BODY_SIZE
+
+    def _add_title(self, doc: Document, room_data: Dict[str, Any]):
+        """Add document title"""
+        title = doc.add_heading(level=0)
+        run = title.add_run(f"事件報告：{room_data.get('title', '未命名事件')}")
+        run.font.size = TITLE_SIZE
+        run.font.bold = True
+
+        # Add generation timestamp
+        timestamp = datetime.utcnow().strftime("%Y-%m-%d %H:%M")
+        subtitle = doc.add_paragraph()
+        subtitle.alignment = WD_ALIGN_PARAGRAPH.CENTER
+        run = subtitle.add_run(f"報告產生時間：{timestamp}")
+        run.font.size = Pt(10)
+        run.font.color.rgb = RGBColor(128, 128, 128)
+
+        doc.add_paragraph()  # Spacing
+
+    def _add_metadata_table(self, doc: Document, room_data: Dict[str, Any]):
+        """Add metadata summary table"""
+        table = doc.add_table(rows=4, cols=4)
+        table.style = "Table Grid"
+
+        # Row 1: Type and Severity
+        cells = table.rows[0].cells
+        cells[0].text = "事件類型"
+        incident_type = room_data.get("incident_type", "other")
+        cells[1].text = self.INCIDENT_TYPE_MAP.get(incident_type, incident_type)
+        cells[2].text = "嚴重程度"
+        severity = room_data.get("severity", "medium")
+        cells[3].text = self.SEVERITY_MAP.get(severity, severity)
+
+        # Row 2: Status and Location
+        cells = table.rows[1].cells
+        cells[0].text = "目前狀態"
+        status = room_data.get("status", "active")
+        cells[1].text = self.STATUS_MAP.get(status, status)
+        cells[2].text = "發生地點"
+        cells[3].text = room_data.get("location") or "未指定"
+
+        # Row 3: Created and Resolved times
+        cells = table.rows[2].cells
+        cells[0].text = "建立時間"
+        created_at = room_data.get("created_at")
+        if isinstance(created_at, datetime):
+            cells[1].text = created_at.strftime("%Y-%m-%d %H:%M")
+        else:
+            cells[1].text = str(created_at) if created_at else "未知"
+
+        cells[2].text = "解決時間"
+        resolved_at = room_data.get("resolved_at")
+        if isinstance(resolved_at, datetime):
+            cells[3].text = resolved_at.strftime("%Y-%m-%d %H:%M")
+        elif resolved_at:
+            cells[3].text = str(resolved_at)
+        else:
+            cells[3].text = "尚未解決"
+
+        # Row 4: Description (spanning all columns)
+        cells = table.rows[3].cells
+        cells[0].text = "事件描述"
+        # Merge remaining cells for description
+        cells[1].merge(cells[3])
+        cells[1].text = room_data.get("description") or "無描述"
+
+        # Style table cells
+        for row in table.rows:
+            for i, cell in enumerate(row.cells):
+                for paragraph in cell.paragraphs:
+                    for run in paragraph.runs:
+                        run.font.size = BODY_SIZE
+                        if i % 2 == 0:  # Header cells
+                            run.font.bold = True
+
+        doc.add_paragraph()  # Spacing
+
+    def _add_summary_section(self, doc: Document, ai_content: Dict[str, Any]):
+        """Add event summary section"""
+        doc.add_heading("事件摘要", level=1)
+        summary = ai_content.get("summary", {})
+        content = summary.get("content", "無摘要內容")
+        p = doc.add_paragraph(content)
+        p.paragraph_format.first_line_indent = Pt(24)
+
+    def _add_timeline_section(self, doc: Document, ai_content: Dict[str, Any]):
+        """Add timeline section"""
+        doc.add_heading("事件時間軸", level=1)
+        timeline = ai_content.get("timeline", {})
+        events = timeline.get("events", [])
+
+        if not events:
+            doc.add_paragraph("無時間軸記錄")
+            return
+
+        # Create timeline table
+        table = doc.add_table(rows=len(events) + 1, cols=2)
+        table.style = "Table Grid"
+
+        # Header row
+        header = table.rows[0].cells
+        header[0].text = "時間"
+        header[1].text = "事件"
+        for cell in header:
+            for run in cell.paragraphs[0].runs:
+                run.font.bold = True
+
+        # Event rows
+        for i, event in enumerate(events):
+            row = table.rows[i + 1].cells
+            row[0].text = event.get("time", "")
+            row[1].text = event.get("description", "")
+
+        doc.add_paragraph()  # Spacing
+
+    def _add_participants_section(self, doc: Document, ai_content: Dict[str, Any]):
+        """Add participants section"""
+        doc.add_heading("參與人員", level=1)
+        participants = ai_content.get("participants", {})
+        members = participants.get("members", [])
+
+        if not members:
+            doc.add_paragraph("無參與人員記錄")
+            return
+
+        # Create participants table
+        table = doc.add_table(rows=len(members) + 1, cols=2)
+        table.style = "Table Grid"
+
+        # Header row
+        header = table.rows[0].cells
+        header[0].text = "姓名"
+        header[1].text = "角色"
+        for cell in header:
+            for run in cell.paragraphs[0].runs:
+                run.font.bold = True
+
+        # Member rows
+        for i, member in enumerate(members):
+            row = table.rows[i + 1].cells
+            row[0].text = member.get("name", "")
+            row[1].text = member.get("role", "")
+
+        doc.add_paragraph()  # Spacing
+
+    def _add_resolution_process_section(self, doc: Document, ai_content: Dict[str, Any]):
+        """Add resolution process section"""
+        doc.add_heading("處理過程", level=1)
+        resolution = ai_content.get("resolution_process", {})
+        content = resolution.get("content", "無處理過程記錄")
+        p = doc.add_paragraph(content)
+        p.paragraph_format.first_line_indent = Pt(24)
+
+    def _add_current_status_section(self, doc: Document, ai_content: Dict[str, Any]):
+        """Add current status section"""
+        doc.add_heading("目前狀態", level=1)
+        current_status = ai_content.get("current_status", {})
+        status = current_status.get("status", "unknown")
+        status_text = self.STATUS_MAP.get(status, status)
+        description = current_status.get("description", "")
+
+        p = doc.add_paragraph()
+        p.add_run(f"狀態：").bold = True
+        p.add_run(status_text)
+
+        if description:
+            doc.add_paragraph(description)
+
+    def _add_final_resolution_section(self, doc: Document, ai_content: Dict[str, Any]):
+        """Add final resolution section"""
+        doc.add_heading("最終處置結果", level=1)
+        final = ai_content.get("final_resolution", {})
+        has_resolution = final.get("has_resolution", False)
+        content = final.get("content", "")
+
+        if has_resolution:
+            if content:
+                p = doc.add_paragraph(content)
+                p.paragraph_format.first_line_indent = Pt(24)
+            else:
+                doc.add_paragraph("事件已解決，但無詳細說明。")
+        else:
+            doc.add_paragraph("事件尚未解決或無最終處置結果。")
+
+    def _add_images_section(self, doc: Document, files: List[Dict[str, Any]]):
+        """Add images section with embedded images from MinIO"""
+        image_files = [f for f in files if f.get("file_type") == "image"]
+
+        if not image_files:
+            return
+
+        doc.add_heading("相關圖片", level=1)
+
+        for f in image_files:
+            try:
+                # Download image from MinIO
+                image_data = self._download_file(f.get("minio_object_path", ""))
+                if image_data:
+                    # Add image to document
+                    doc.add_picture(image_data, width=Inches(5))
+
+                    # Add caption
+                    caption = doc.add_paragraph()
+                    caption.alignment = WD_ALIGN_PARAGRAPH.CENTER
+                    run = caption.add_run(f"{f.get('filename', '圖片')}")
+                    run.font.size = Pt(9)
+                    run.font.italic = True
+
+                    doc.add_paragraph()  # Spacing
+                else:
+                    # Image download failed, add note
+                    doc.add_paragraph(f"[圖片載入失敗: {f.get('filename', '未知')}]")
+
+            except Exception as e:
+                logger.error(f"Failed to embed image {f.get('filename')}: {e}")
+                doc.add_paragraph(f"[圖片嵌入失敗: {f.get('filename', '未知')}]")
+
+    def _add_file_list_section(self, doc: Document, files: List[Dict[str, Any]]):
+        """Add file attachment list section"""
+        doc.add_heading("附件清單", level=1)
+
+        if not files:
+            doc.add_paragraph("無附件")
+            return
+
+        # Create file list table
+        table = doc.add_table(rows=len(files) + 1, cols=4)
+        table.style = "Table Grid"
+
+        # Header row
+        header = table.rows[0].cells
+        header[0].text = "檔案名稱"
+        header[1].text = "類型"
+        header[2].text = "上傳者"
+        header[3].text = "上傳時間"
+        for cell in header:
+            for run in cell.paragraphs[0].runs:
+                run.font.bold = True
+
+        # File type mapping
+        file_type_map = {
+            "image": "圖片",
+            "document": "文件",
+            "log": "記錄檔",
+        }
+
+        # File rows
+        for i, f in enumerate(files):
+            row = table.rows[i + 1].cells
+            row[0].text = f.get("filename", "")
+            file_type = f.get("file_type", "file")
+            row[1].text = file_type_map.get(file_type, file_type)
+            row[2].text = f.get("uploader_name") or f.get("uploader_id", "")
+
+            uploaded_at = f.get("uploaded_at")
+            if isinstance(uploaded_at, datetime):
+                row[3].text = uploaded_at.strftime("%Y-%m-%d %H:%M")
+            else:
+                row[3].text = str(uploaded_at) if uploaded_at else ""
+
+    def _download_file(self, object_path: str) -> Optional[io.BytesIO]:
+        """Download file from MinIO
+
+        Args:
+            object_path: MinIO object path
+
+        Returns:
+            BytesIO containing file data, or None if download fails
+        """
+        if not object_path:
+            return None
+
+        try:
+            response = self.minio_client.get_object(self.bucket, object_path)
+            data = io.BytesIO(response.read())
+            response.close()
+            response.release_conn()
+            return data
+        except Exception as e:
+            logger.error(f"Failed to download file from MinIO: {object_path} - {e}")
+            return None
+
+    def upload_report(
+        self,
+        report_data: io.BytesIO,
+        room_id: str,
+        report_id: str,
+    ) -> Optional[str]:
+        """Upload generated report to MinIO
+
+        Args:
+            report_data: BytesIO containing the .docx file
+            room_id: Room ID for path organization
+            report_id: Report ID for unique filename
+
+        Returns:
+            MinIO object path if successful, None otherwise
+        """
+        timestamp = datetime.utcnow().strftime("%Y%m%d_%H%M%S")
+        object_path = f"{settings.REPORT_STORAGE_PATH}/{room_id}/{report_id}_{timestamp}.docx"
+
+        # Get file size
+        report_data.seek(0, 2)  # Seek to end
+        file_size = report_data.tell()
+        report_data.seek(0)  # Reset to beginning
+
+        success = upload_file(
+            bucket=self.bucket,
+            object_path=object_path,
+            file_data=report_data,
+            file_size=file_size,
+            content_type="application/vnd.openxmlformats-officedocument.wordprocessingml.document",
+        )
+
+        if success:
+            logger.info(f"Report uploaded to MinIO: {object_path}")
+            return object_path
+        else:
+            logger.error(f"Failed to upload report to MinIO")
+            return None
+
+
+# Singleton instance
+docx_service = DocxAssemblyService()

app/modules/report_generation/services/report_data_service.py

@@ -0,0 +1,263 @@
+"""Report Data Collection Service
+
+Collects all room data needed for AI report generation:
+- Room metadata (title, type, severity, status, etc.)
+- Messages with sender display names
+- Room members with display names
+- File attachments with uploader names
+"""
+from typing import Dict, Any, List, Optional
+from dataclasses import dataclass, field
+from datetime import datetime
+from sqlalchemy.orm import Session
+from sqlalchemy import desc
+
+from app.modules.chat_room.models import IncidentRoom, RoomMember
+from app.modules.realtime.models import Message
+from app.modules.file_storage.models import RoomFile
+from app.modules.auth.models import User
+
+
+@dataclass
+class MessageData:
+    """Message data for report generation"""
+    message_id: str
+    sender_id: str
+    sender_name: str
+    content: str
+    message_type: str
+    created_at: datetime
+    file_name: Optional[str] = None
+
+
+@dataclass
+class MemberData:
+    """Member data for report generation"""
+    user_id: str
+    display_name: str
+    role: str
+
+
+@dataclass
+class FileData:
+    """File data for report generation"""
+    file_id: str
+    filename: str
+    file_type: str
+    mime_type: str
+    uploaded_at: datetime
+    uploader_id: str
+    uploader_name: str
+    minio_object_path: str
+
+
+@dataclass
+class RoomReportData:
+    """Complete room data for report generation"""
+    room_id: str
+    title: str
+    incident_type: str
+    severity: str
+    status: str
+    location: Optional[str]
+    description: Optional[str]
+    resolution_notes: Optional[str]
+    created_at: datetime
+    resolved_at: Optional[datetime]
+    created_by: str
+    messages: List[MessageData] = field(default_factory=list)
+    members: List[MemberData] = field(default_factory=list)
+    files: List[FileData] = field(default_factory=list)
+
+
+class ReportDataService:
+    """Service to collect room data for report generation"""
+
+    def __init__(self, db: Session):
+        self.db = db
+
+    def collect_room_data(self, room_id: str) -> Optional[RoomReportData]:
+        """Collect all data for a room
+
+        Args:
+            room_id: Room ID to collect data for
+
+        Returns:
+            RoomReportData with all collected data, or None if room not found
+        """
+        # Get room metadata
+        room = self.db.query(IncidentRoom).filter(
+            IncidentRoom.room_id == room_id
+        ).first()
+
+        if not room:
+            return None
+
+        # Collect messages with sender names
+        messages = self._collect_messages(room_id)
+
+        # Collect members with display names
+        members = self._collect_members(room_id)
+
+        # Collect files with uploader names
+        files = self._collect_files(room_id)
+
+        return RoomReportData(
+            room_id=room.room_id,
+            title=room.title,
+            incident_type=room.incident_type.value if room.incident_type else "other",
+            severity=room.severity.value if room.severity else "medium",
+            status=room.status.value if room.status else "active",
+            location=room.location,
+            description=room.description,
+            resolution_notes=room.resolution_notes,
+            created_at=room.created_at,
+            resolved_at=room.resolved_at,
+            created_by=room.created_by,
+            messages=messages,
+            members=members,
+            files=files,
+        )
+
+    def _collect_messages(self, room_id: str) -> List[MessageData]:
+        """Collect messages with sender display names"""
+        # Query messages with LEFT JOIN to users table for display names
+        results = (
+            self.db.query(Message, User.display_name)
+            .outerjoin(User, Message.sender_id == User.user_id)
+            .filter(Message.room_id == room_id)
+            .filter(Message.deleted_at.is_(None))  # Exclude deleted messages
+            .order_by(Message.created_at)
+            .all()
+        )
+
+        messages = []
+        for msg, display_name in results:
+            # Extract file name from metadata if it's a file reference
+            file_name = None
+            if msg.message_type.value in ("image_ref", "file_ref") and msg.message_metadata:
+                file_name = msg.message_metadata.get("filename")
+
+            messages.append(MessageData(
+                message_id=msg.message_id,
+                sender_id=msg.sender_id,
+                sender_name=display_name or msg.sender_id,  # Fallback to sender_id
+                content=msg.content,
+                message_type=msg.message_type.value,
+                created_at=msg.created_at,
+                file_name=file_name,
+            ))
+
+        return messages
+
+    def _collect_members(self, room_id: str) -> List[MemberData]:
+        """Collect room members with display names"""
+        results = (
+            self.db.query(RoomMember, User.display_name)
+            .outerjoin(User, RoomMember.user_id == User.user_id)
+            .filter(RoomMember.room_id == room_id)
+            .filter(RoomMember.removed_at.is_(None))  # Exclude removed members
+            .all()
+        )
+
+        members = []
+        for member, display_name in results:
+            members.append(MemberData(
+                user_id=member.user_id,
+                display_name=display_name or member.user_id,  # Fallback to user_id
+                role=member.role.value if member.role else "viewer",
+            ))
+
+        return members
+
+    def _collect_files(self, room_id: str) -> List[FileData]:
+        """Collect room files with uploader display names"""
+        results = (
+            self.db.query(RoomFile, User.display_name)
+            .outerjoin(User, RoomFile.uploader_id == User.user_id)
+            .filter(RoomFile.room_id == room_id)
+            .filter(RoomFile.deleted_at.is_(None))  # Exclude deleted files
+            .order_by(RoomFile.uploaded_at)
+            .all()
+        )
+
+        files = []
+        for f, display_name in results:
+            files.append(FileData(
+                file_id=f.file_id,
+                filename=f.filename,
+                file_type=f.file_type,
+                mime_type=f.mime_type,
+                uploaded_at=f.uploaded_at,
+                uploader_id=f.uploader_id,
+                uploader_name=display_name or f.uploader_id,  # Fallback to uploader_id
+                minio_object_path=f.minio_object_path,
+            ))
+
+        return files
+
+    def to_prompt_dict(self, data: RoomReportData) -> Dict[str, Any]:
+        """Convert RoomReportData to dictionary format for prompt builder
+
+        Args:
+            data: RoomReportData object
+
+        Returns:
+            Dictionary with room_data, messages, members, files keys
+        """
+        room_data = {
+            "room_id": data.room_id,
+            "title": data.title,
+            "incident_type": data.incident_type,
+            "severity": data.severity,
+            "status": data.status,
+            "location": data.location,
+            "description": data.description,
+            "resolution_notes": data.resolution_notes,
+            "created_at": data.created_at,
+            "resolved_at": data.resolved_at,
+            "created_by": data.created_by,
+        }
+
+        messages = [
+            {
+                "message_id": m.message_id,
+                "sender_id": m.sender_id,
+                "sender_name": m.sender_name,
+                "content": m.content,
+                "message_type": m.message_type,
+                "created_at": m.created_at,
+                "file_name": m.file_name,
+            }
+            for m in data.messages
+        ]
+
+        members = [
+            {
+                "user_id": m.user_id,
+                "display_name": m.display_name,
+                "role": m.role,
+            }
+            for m in data.members
+        ]
+
+        files = [
+            {
+                "file_id": f.file_id,
+                "filename": f.filename,
+                "file_type": f.file_type,
+                "mime_type": f.mime_type,
+                "uploaded_at": f.uploaded_at,
+                "uploader_id": f.uploader_id,
+                "uploader_name": f.uploader_name,
+                "minio_object_path": f.minio_object_path,
+            }
+            for f in data.files
+        ]
+
+        return {
+            "room_data": room_data,
+            "messages": messages,
+            "members": members,
+            "files": files,
+        }