diff --git a/backend/app/api/router.py b/backend/app/api/router.py
index 84c67ee..dd46043 100644
--- a/backend/app/api/router.py
+++ b/backend/app/api/router.py
@@ -11,4 +11,4 @@ api_router = APIRouter(prefix=settings.API_V1_PREFIX)
 api_router.include_router(models_router)
 api_router.include_router(templates_router)
 api_router.include_router(generation_points_router)
-api_router.include_router(tasks_router)
+api_router.include_router(tasks_router, prefix="")
diff --git a/backend/app/api/tasks.py b/backend/app/api/tasks.py
index 4fe82c7..1ecf604 100644
--- a/backend/app/api/tasks.py
+++ b/backend/app/api/tasks.py
@@ -1,5 +1,153 @@
-from fastapi import APIRouter
+import os
+import uuid
+from datetime import datetime, timezone
+from fastapi import APIRouter, Depends, HTTPException, Query
+from fastapi.responses import Response
+from sqlalchemy.ext.asyncio import AsyncSession
+from sqlalchemy import select
+from app.core.database import get_db
+from app.models.generation_task import GenerationTask
+from app.models.generation_point import GenerationPoint
+from app.models.template import Template
+from app.models.ai_model import AIModel
+from app.schemas.generation_task import TaskResponse, GenerateResponse, SingleTestRequest, SingleTestResponse
+from app.services.document_processor import docx_to_pdf_bytes
+from app.services.file_storage import get_file_content
+from app.services.ai_adapter import call_ai_model
+from app.services.ref_parser import parse_reference_file
+from app.tasks.generate import generate_document
 
-router = APIRouter(prefix="/tasks", tags=["生成任务"])
+router = APIRouter()
 
-# 任务管理 API 将在阶段三实现
+
+@router.post("/templates/{template_id}/generate", response_model=GenerateResponse)
+async def trigger_generation(template_id: str, db: AsyncSession = Depends(get_db)):
+    template_result = await db.execute(select(Template).where(Template.id == template_id))
+    if not template_result.scalar_one_or_none():
+        raise HTTPException(status_code=404, detail="模板不存在")
+
+    points_result = await db.execute(
+        select(GenerationPoint).where(GenerationPoint.template_id == template_id)
+    )
+    points = points_result.scalars().all()
+    if not points:
+        raise HTTPException(status_code=400, detail="模板没有生成点")
+
+    task = GenerationTask(
+        template_id=uuid.UUID(template_id),
+        status="pending",
+        created_at=datetime.now(timezone.utc),
+    )
+    db.add(task)
+    await db.flush()
+
+    celery_task = generate_document.delay(str(task.id))
+    task.celery_task_id = celery_task.id
+    await db.commit()
+
+    return GenerateResponse(task_id=task.id, status="pending")
+
+
+@router.get("/tasks/{task_id}", response_model=TaskResponse)
+async def get_task_status(task_id: str, db: AsyncSession = Depends(get_db)):
+    result = await db.execute(select(GenerationTask).where(GenerationTask.id == task_id))
+    task = result.scalar_one_or_none()
+    if not task:
+        raise HTTPException(status_code=404, detail="任务不存在")
+    return task
+
+
+@router.get("/tasks/{task_id}/download")
+async def download_task_result(
+    task_id: str,
+    format: str = Query("docx", pattern="^(docx|pdf)$"),
+    db: AsyncSession = Depends(get_db),
+):
+    result = await db.execute(select(GenerationTask).where(GenerationTask.id == task_id))
+    task = result.scalar_one_or_none()
+    if not task:
+        raise HTTPException(status_code=404, detail="任务不存在")
+
+    if task.status != "done" or not task.result_file_path:
+        raise HTTPException(status_code=400, detail="任务未完成或无结果文件")
+
+    file_content = await get_file_content(task.result_file_path)
+
+    if format == "pdf":
+        file_content = docx_to_pdf_bytes(file_content)
+        media_type = "application/pdf"
+        filename = f"result_{task_id}.pdf"
+    else:
+        media_type = "application/vnd.openxmlformats-officedocument.wordprocessingml.document"
+        filename = f"result_{task_id}.docx"
+
+    return Response(
+        content=file_content,
+        media_type=media_type,
+        headers={"Content-Disposition": f'attachment; filename="{filename}"'},
+    )
+
+
+@router.get("/tasks")
+async def list_tasks(
+    skip: int = Query(0, ge=0),
+    limit: int = Query(20, ge=1, le=100),
+    db: AsyncSession = Depends(get_db),
+):
+    query = select(GenerationTask).offset(skip).limit(limit).order_by(GenerationTask.created_at.desc())
+    result = await db.execute(query)
+    return result.scalars().all()
+
+
+@router.post("/tasks/{task_id}/cancel")
+async def cancel_task(task_id: str, db: AsyncSession = Depends(get_db)):
+    result = await db.execute(select(GenerationTask).where(GenerationTask.id == task_id))
+    task = result.scalar_one_or_none()
+    if not task:
+        raise HTTPException(status_code=404, detail="任务不存在")
+
+    if task.status not in ("pending", "processing"):
+        raise HTTPException(status_code=400, detail="任务无法取消")
+
+    if task.celery_task_id:
+        from app.tasks.celery_app import celery_app
+        celery_app.control.revoke(task.celery_task_id, terminate=True)
+
+    task.status = "failed"
+    task.error_msg = "用户取消"
+    task.finished_at = datetime.now(timezone.utc)
+    await db.commit()
+    return {"detail": "任务已取消"}
+
+
+@router.post("/generation-points/{point_id}/test", response_model=SingleTestResponse)
+async def test_single_point(
+    point_id: str,
+    db: AsyncSession = Depends(get_db),
+):
+    point_result = await db.execute(select(GenerationPoint).where(GenerationPoint.id == point_id))
+    point = point_result.scalar_one_or_none()
+    if not point:
+        raise HTTPException(status_code=404, detail="生成点不存在")
+
+    model_config = {"provider": "custom", "endpoint": "", "api_key": "", "extra_params": {}}
+    if point.model_id:
+        model_result = await db.execute(select(AIModel).where(AIModel.id == point.model_id))
+        model = model_result.scalar_one_or_none()
+        if model:
+            model_config = {
+                "provider": model.provider,
+                "endpoint": model.endpoint,
+                "api_key": model.api_key,
+                "extra_params": model.extra_params,
+            }
+
+    ref_content = None
+    if point.ref_file_path:
+        ref_content = await parse_reference_file(point.ref_file_path)
+
+    try:
+        ai_result = await call_ai_model(model_config, point.prompt, ref_content)
+        return SingleTestResponse(result=ai_result)
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=f"AI 调用失败: {str(e)}")
diff --git a/backend/app/services/ai_adapter.py b/backend/app/services/ai_adapter.py
new file mode 100644
index 0000000..2e8cf38
--- /dev/null
+++ b/backend/app/services/ai_adapter.py
@@ -0,0 +1,163 @@
+import json
+from abc import ABC, abstractmethod
+import httpx
+from app.core.security import decrypt_api_key
+
+PROVIDER_OPENAI = "openai"
+PROVIDER_AZURE = "azure"
+PROVIDER_CUSTOM = "custom"
+
+
+class AIAdapter(ABC):
+    @abstractmethod
+    def build_request(self, model_config: dict, prompt: str, ref_content: str | None = None) -> dict:
+        pass
+
+    @abstractmethod
+    def parse_response(self, response_data: dict) -> str:
+        pass
+
+    @property
+    @abstractmethod
+    def provider(self) -> str:
+        pass
+
+
+class OpenAIAdapter(AIAdapter):
+    @property
+    def provider(self) -> str:
+        return PROVIDER_OPENAI
+
+    def build_request(self, model_config: dict, prompt: str, ref_content: str | None = None) -> dict:
+        extra = model_config.get("extra_params", {})
+        temperature = extra.get("temperature", 0.7)
+        max_tokens = extra.get("max_tokens", 2000)
+
+        messages = [{"role": "system", "content": "你是一个专业的文档内容生成助手。"}]
+        user_content = prompt
+        if ref_content:
+            user_content = f"参考以下内容：\n{ref_content}\n\n任务：{prompt}"
+        messages.append({"role": "user", "content": user_content})
+
+        return {
+            "url": model_config["endpoint"],
+            "headers": {
+                "Authorization": f"Bearer {decrypt_api_key(model_config['api_key'])}",
+                "Content-Type": "application/json",
+            },
+            "json": {
+                "model": extra.get("model", "gpt-4"),
+                "messages": messages,
+                "temperature": temperature,
+                "max_tokens": max_tokens,
+            },
+        }
+
+    def parse_response(self, response_data: dict) -> str:
+        return response_data["choices"][0]["message"]["content"]
+
+
+class AzureAdapter(AIAdapter):
+    @property
+    def provider(self) -> str:
+        return PROVIDER_AZURE
+
+    def build_request(self, model_config: dict, prompt: str, ref_content: str | None = None) -> dict:
+        extra = model_config.get("extra_params", {})
+        temperature = extra.get("temperature", 0.7)
+        max_tokens = extra.get("max_tokens", 2000)
+
+        messages = [{"role": "system", "content": "你是一个专业的文档内容生成助手。"}]
+        user_content = prompt
+        if ref_content:
+            user_content = f"参考以下内容：\n{ref_content}\n\n任务：{prompt}"
+        messages.append({"role": "user", "content": user_content})
+
+        api_version = extra.get("api_version", "2024-02-15-preview")
+        endpoint = model_config["endpoint"]
+        url = f"{endpoint}?api-version={api_version}"
+
+        return {
+            "url": url,
+            "headers": {
+                "api-key": decrypt_api_key(model_config["api_key"]),
+                "Content-Type": "application/json",
+            },
+            "json": {
+                "messages": messages,
+                "temperature": temperature,
+                "max_tokens": max_tokens,
+            },
+        }
+
+    def parse_response(self, response_data: dict) -> str:
+        return response_data["choices"][0]["message"]["content"]
+
+
+class CustomAdapter(AIAdapter):
+    @property
+    def provider(self) -> str:
+        return PROVIDER_CUSTOM
+
+    def build_request(self, model_config: dict, prompt: str, ref_content: str | None = None) -> dict:
+        extra = model_config.get("extra_params", {})
+        user_content = prompt
+        if ref_content:
+            user_content = f"参考以下内容：\n{ref_content}\n\n任务：{prompt}"
+
+        body = {
+            "prompt": user_content,
+            "max_tokens": extra.get("max_tokens", 2000),
+            "temperature": extra.get("temperature", 0.7),
+        }
+        body.update({k: v for k, v in extra.items() if k not in ("max_tokens", "temperature")})
+
+        return {
+            "url": model_config["endpoint"],
+            "headers": {
+                "Authorization": f"Bearer {decrypt_api_key(model_config['api_key'])}",
+                "Content-Type": "application/json",
+            },
+            "json": body,
+        }
+
+    def parse_response(self, response_data: dict) -> str:
+        if "choices" in response_data:
+            return response_data["choices"][0]["message"]["content"]
+        if "response" in response_data:
+            return response_data["response"]
+        if "content" in response_data:
+            return response_data["content"]
+        if "text" in response_data:
+            return response_data["text"]
+        return json.dumps(response_data)
+
+
+_adapters: dict[str, AIAdapter] = {
+    PROVIDER_OPENAI: OpenAIAdapter(),
+    PROVIDER_AZURE: AzureAdapter(),
+    PROVIDER_CUSTOM: CustomAdapter(),
+}
+
+
+def get_adapter(provider: str) -> AIAdapter:
+    adapter = _adapters.get(provider)
+    if not adapter:
+        raise ValueError(f"不支持的供应商: {provider}")
+    return adapter
+
+
+async def call_ai_model(model_config: dict, prompt: str, ref_content: str | None = None) -> str:
+    adapter = get_adapter(model_config["provider"])
+    request = adapter.build_request(model_config, prompt, ref_content)
+
+    timeout = model_config.get("extra_params", {}).get("timeout", 120)
+
+    async with httpx.AsyncClient(timeout=timeout) as client:
+        response = await client.post(
+            request["url"],
+            headers=request["headers"],
+            json=request["json"],
+        )
+        response.raise_for_status()
+        return adapter.parse_response(response.json())
diff --git a/backend/app/services/ref_parser.py b/backend/app/services/ref_parser.py
new file mode 100644
index 0000000..4488114
--- /dev/null
+++ b/backend/app/services/ref_parser.py
@@ -0,0 +1,25 @@
+import os
+from docx import Document
+from PyPDF2 import PdfReader
+from app.services.file_storage import get_file_content
+
+
+async def parse_reference_file(file_path: str) -> str:
+    ext = os.path.splitext(file_path)[1].lower()
+
+    content = await get_file_content(file_path)
+
+    if ext == ".txt":
+        return content.decode("utf-8", errors="ignore")
+
+    if ext == ".docx":
+        from io import BytesIO
+        doc = Document(BytesIO(content))
+        return "\n".join(p.text for p in doc.paragraphs if p.text.strip())
+
+    if ext == ".pdf":
+        from io import BytesIO
+        reader = PdfReader(BytesIO(content))
+        return "\n".join(page.extract_text() or "" for page in reader.pages)
+
+    raise ValueError(f"不支持的文件格式: {ext}")
diff --git a/backend/app/tasks/__init__.py b/backend/app/tasks/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/backend/app/tasks/celery_app.py b/backend/app/tasks/celery_app.py
new file mode 100644
index 0000000..76f876f
--- /dev/null
+++ b/backend/app/tasks/celery_app.py
@@ -0,0 +1,25 @@
+from celery import Celery
+from app.core.config import get_settings
+
+settings = get_settings()
+
+celery_app = Celery(
+    "doc_forge_reds",
+    broker=settings.CELERY_BROKER_URL,
+    backend=settings.CELERY_RESULT_BACKEND,
+)
+
+celery_app.conf.update(
+    task_serializer="json",
+    accept_content=["json"],
+    result_serializer="json",
+    timezone="Asia/Shanghai",
+    enable_utc=True,
+    task_track_started=True,
+    task_acks_late=True,
+    worker_prefetch_multiplier=1,
+    task_soft_time_limit=600,
+    task_time_limit=900,
+)
+
+celery_app.autodiscover_tasks(["app.tasks.generate"])
diff --git a/backend/app/tasks/generate.py b/backend/app/tasks/generate.py
new file mode 100644
index 0000000..b474c3e
--- /dev/null
+++ b/backend/app/tasks/generate.py
@@ -0,0 +1,107 @@
+import os
+import uuid
+from datetime import datetime, timezone
+from sqlalchemy import select
+from app.tasks.celery_app import celery_app
+from app.core.database import async_session_factory
+from app.core.config import get_settings
+from app.models.generation_task import GenerationTask
+from app.models.generation_point import GenerationPoint
+from app.models.template import Template
+from app.services.document_processor import html_to_docx_bytes
+from app.services.ai_adapter import call_ai_model
+from app.services.ref_parser import parse_reference_file
+from app.services.file_storage import get_storage_dir, RESULTS_DIR
+
+
+async def _generate_document(task_id: str) -> None:
+    settings = get_settings()
+
+    async with async_session_factory() as db:
+        task_result = await db.execute(select(GenerationTask).where(GenerationTask.id == task_id))
+        task = task_result.scalar_one_or_none()
+        if not task:
+            return
+
+        task.status = "processing"
+        await db.commit()
+
+        template_result = await db.execute(select(Template).where(Template.id == task.template_id))
+        template = template_result.scalar_one_or_none()
+        if not template:
+            task.status = "failed"
+            task.error_msg = "模板不存在"
+            task.finished_at = datetime.now(timezone.utc)
+            await db.commit()
+            return
+
+        html_content = template.html_content or ""
+
+        points_result = await db.execute(
+            select(GenerationPoint)
+            .where(GenerationPoint.template_id == task.template_id)
+            .order_by(GenerationPoint.order.asc())
+        )
+        points = points_result.scalars().all()
+
+        try:
+            total = len(points)
+            for idx, point in enumerate(points):
+                ref_content = None
+                if point.ref_file_path:
+                    ref_content = await parse_reference_file(point.ref_file_path)
+
+                model_config = {"provider": "custom", "endpoint": "", "api_key": "", "extra_params": {}}
+                if point.model_id:
+                    from app.models.ai_model import AIModel
+                    model_result = await db.execute(select(AIModel).where(AIModel.id == point.model_id))
+                    model = model_result.scalar_one_or_none()
+                    if model:
+                        model_config = {
+                            "provider": model.provider,
+                            "endpoint": model.endpoint,
+                            "api_key": model.api_key,
+                            "extra_params": model.extra_params,
+                        }
+
+                ai_result = await call_ai_model(model_config, point.prompt, ref_content)
+
+                position = point.position
+                start = position.get("start", 0)
+                end = position.get("end", 0)
+                if 0 <= start < end <= len(html_content):
+                    html_content = html_content[:start] + ai_result + html_content[end:]
+
+                task.status = "processing"
+                await db.commit()
+
+            result_dir = get_storage_dir(RESULTS_DIR)
+            result_path = os.path.join(result_dir, f"{task_id}.docx")
+            docx_bytes = html_to_docx_bytes(html_content)
+            with open(result_path, "wb") as f:
+                f.write(docx_bytes)
+
+            task.status = "done"
+            task.result_file_path = result_path
+            task.finished_at = datetime.now(timezone.utc)
+            await db.commit()
+
+        except Exception as e:
+            task.status = "failed"
+            task.error_msg = str(e)
+            task.finished_at = datetime.now(timezone.utc)
+            await db.commit()
+
+
+@celery_app.task(bind=True, name="generate_document")
+def generate_document(self, task_id: str) -> dict:
+    import asyncio
+    loop = asyncio.new_event_loop()
+    asyncio.set_event_loop(loop)
+    try:
+        loop.run_until_complete(_generate_document(task_id))
+        return {"status": "done", "task_id": task_id}
+    except Exception as e:
+        return {"status": "failed", "task_id": task_id, "error": str(e)}
+    finally:
+        loop.close()
diff --git a/docs/tasks/task_detail_2026_07_06.md b/docs/tasks/task_detail_2026_07_06.md
new file mode 100644
index 0000000..b78af29
--- /dev/null
+++ b/docs/tasks/task_detail_2026_07_06.md
@@ -0,0 +1,24 @@
+# 任务执行摘要
+
+## 会话 ID: 1
+- [2026-07-06 16:00]
+- **执行原因**: 用户要求按任务拆解清单逐步实施项目开发，先完成阶段一和阶段二
+- **执行过程**:
+    1. 创建项目目录结构（backend/, web/, storage/）。
+    2. 编写 docker-compose.yml（Postgres 18.3 + Redis 7）、.env.example、requirements.txt。
+    3. 实现 backend/app/core/ 核心模块：config（pydantic-settings）、database（SQLAlchemy async engine + session）、security（Fernet 加密）。
+    4. 创建 5 个 ORM 模型：AIModel、Template、GenerationPoint、GenerationTask、SystemConfig + Base/TimestampMixin/UUIDMixin。
+    5. 配置 Alembic（alembic.ini, env.py, script.py.mako）。
+    6. 编写 FastAPI 入口 main.py（CORS + health 端点）。
+    7. 实现文件存储服务 file_storage.py（上传/下载/删除）。
+    8. 初始化前端项目（Vite + React 18 + TypeScript + Ant Design + React Router），侧边栏布局 + 5 个占位页面。
+    9. 编写 4 组 Pydantic Schemas（ai_model, template, generation_point, generation_task）。
+    10. 实现文档处理服务 document_processor.py（Mammoth docx→HTML、python-docx HTML→docx、reportlab docx→PDF）。
+    11. 实现 AI 模型管理 API（CRUD + 启用/禁用切换，6 个端点）。
+    12. 实现模板管理 API（上传/获取HTML/更新HTML/下载docx+pdf，6 个端点）。
+    13. 实现生成点管理 API（CRUD + 参考文件上传 + 批量排序更新，5 个端点）。
+    14. 注册路由（api/router.py），更新 main.py，启动验证通过（/health 返回 OK）。
+- **执行结果**: 
+    - 阶段一（基础设施）：45 个文件，后端 5 张表模型导入正常，前端 TypeScript 编译零错误。
+    - 阶段二（核心业务 API）：17 个 API 端点全部注册，后端启动正常，`/health` 返回 200 OK。
+    - Git 提交：2 次提交（b5fa343, aa84521）。