Add FastAPI backend with agent system

backend/app/services/todo_service.py

@@ -0,0 +1,165 @@
+import json
+import logging
+from datetime import date, datetime, timedelta
+from sqlalchemy.ext.asyncio import AsyncSession
+from sqlalchemy import select
+from app.models.todo import DailyTodo, TodoSource
+from app.models.task import Task, TaskStatus
+from app.models.conversation import Conversation, Message
+from app.services.llm_service import get_llm
+from langchain_core.messages import HumanMessage, SystemMessage
+
+logger = logging.getLogger(__name__)
+
+
+async def generate_daily_todos(user_id: str, db: AsyncSession) -> list[DailyTodo]:
+    """
+    为用户生成今日待办：
+    1. 来自前一天未完成的看板任务（最多20条）
+    2. 来自前一天对话记录分析（最多3条）
+    """
+    today = date.today()
+    yesterday = (today - timedelta(days=1)).isoformat()
+
+    todos: list[DailyTodo] = []
+
+    # 1. 从看板任务导入
+    kanban_todos = await _import_kanban_tasks(user_id, yesterday, db)
+    todos.extend(kanban_todos)
+
+    # 2. 从对话记录分析
+    chat_todos = await _analyze_chat_history(user_id, yesterday, db)
+    todos.extend(chat_todos)
+
+    return todos
+
+
+async def _import_kanban_tasks(user_id: str, date_str: str, db: AsyncSession) -> list[DailyTodo]:
+    """导入前一天创建的、未完成的看板任务"""
+    q = select(Task).where(
+        Task.user_id == user_id,
+        Task.status != TaskStatus.DONE,
+    ).order_by(Task.created_at.desc()).limit(20)
+
+    tasks = (await db.execute(q)).scalars().all()
+    todos = []
+
+    for task in tasks:
+        todo = DailyTodo(
+            user_id=user_id,
+            title=task.title,
+            source=TodoSource.AI_KANBAN,
+            source_detail=f"看板：{task.title}",
+            source_ref_id=task.id,
+            todo_date=date.today().isoformat(),
+        )
+        db.add(todo)
+        todos.append(todo)
+
+    if todos:
+        await db.commit()
+        for todo in todos:
+            await db.refresh(todo)
+
+    return todos
+
+
+async def _analyze_chat_history(user_id: str, date_str: str, db: AsyncSession) -> list[DailyTodo]:
+    """分析前一天对话，提取待办事项"""
+    try:
+        # 查询前一天创建的对话
+        conv_q = select(Conversation).where(
+            Conversation.user_id == user_id,
+        ).order_by(Conversation.created_at.desc()).limit(10)
+        convs = (await db.execute(conv_q)).scalars().all()
+
+        # 过滤出昨天的对话
+        yesterday_convs = []
+        for conv in convs:
+            created = conv.created_at
+            if hasattr(created, 'date'):
+                created_date = created.date() if hasattr(created, 'date') else created
+            else:
+                created_date = datetime.fromisoformat(str(created)).date()
+
+            if str(created_date) == date_str or (created + timedelta(hours=8)).strftime('%Y-%m-%d') == date_str:
+                yesterday_convs.append(conv)
+
+        if not yesterday_convs:
+            return []
+
+        # 收集消息内容（限制2000字）
+        messages_content = []
+        for conv in yesterday_convs:
+            msg_q = select(Message).where(
+                Message.conversation_id == conv.id
+            ).order_by(Message.created_at.asc()).limit(50)
+            msgs = (await db.execute(msg_q)).scalars().all()
+            for msg in msgs:
+                if msg.content:
+                    messages_content.append(f"[{msg.role}]: {msg.content[:500]}")
+
+        if not messages_content:
+            return []
+
+        full_text = "\n".join(messages_content)[:2000]
+
+        # 调用 LLM 分析
+        prompt = f"""你是一个任务规划助手。请分析以下对话记录，提取其中用户想要完成但尚未明确完成的事项。
+
+要求：
+- 最多提取 3 条
+- 每条格式：{{"title": "事项描述（50字以内）", "reason": "来源说明（60字以内）"}}
+- 只提取用户明确表达过需求但还未完成的事项
+- 如果没有可提取的内容，返回空数组 []
+
+对话记录：
+{full_text}
+
+返回 JSON 数组："""
+
+        llm = get_llm()
+        response = await llm.invoke([
+            SystemMessage(content="你是一个任务规划助手。"),
+            HumanMessage(content=prompt),
+        ])
+        content = response.content if hasattr(response, 'content') else str(response)
+
+        # 尝试解析 JSON
+        try:
+            # 提取 JSON 数组
+            start = content.find('[')
+            end = content.rfind(']') + 1
+            if start != -1 and end > start:
+                items = json.loads(content[start:end])
+            else:
+                items = []
+        except (json.JSONDecodeError, ValueError):
+            logger.warning(f"LLM 返回格式异常，跳过对话分析: {content[:200]}")
+            items = []
+
+        if not items:
+            return []
+
+        todos = []
+        for item in items[:3]:
+            todo = DailyTodo(
+                user_id=user_id,
+                title=item.get("title", "")[:500],
+                source=TodoSource.AI_CHAT,
+                source_detail=f"对话：{item.get('reason', '')[:60]}",
+                todo_date=date.today().isoformat(),
+            )
+            db.add(todo)
+            todos.append(todo)
+
+        if todos:
+            await db.commit()
+            for todo in todos:
+                await db.refresh(todo)
+
+        return todos
+
+    except Exception as e:
+        logger.error(f"对话分析失败: {e}")
+        return []