fix: 核心 bug 修复 - API Key 加密/解密 + AI 调用 + 数据安全

- schema: 用 @field_serializer 掩码 api_key，不再直接修改 SQLAlchemy 对象
- API: 移除所有 model.api_key='***' 赋值，防止数据库被覆盖
- security: SHA256(SECRET_KEY) 派生合法 Fernet 32 字节密钥
- ai_adapter: CustomAdapter 改用 OpenAI-compatible messages 格式
- ai_adapter: trust_env=False 防止代理干扰，增加详细异常信息
- templates: bleach HTML 净化防止 XSS
- .env.example: CORS_ORIGINS 修复为 JSON 数组格式
- 端到端验证：单点测试 + 异步生成任务均通过

.env.example

@@ -2,9 +2,25 @@
 POSTGRES_USER=docforge
 POSTGRES_PASSWORD=change-this-password
 POSTGRES_DB=docforge
+POSTGRES_HOST=localhost
+POSTGRES_PORT=5432
 
-# Security (generate with: python -c "import secrets; print(secrets.token_urlsafe(32))")
+# Redis
+REDIS_URL=redis://localhost:6379/0
+
+# Backend
 SECRET_KEY=change-me-to-a-random-secret-key
+DEFAULT_MODEL_ID=
+
+# File Storage
+STORAGE_ROOT=./storage
+
+# Celery
+CELERY_BROKER_URL=redis://localhost:6379/0
+CELERY_RESULT_BACKEND=redis://localhost:6379/1
+
+# CORS (JSON array format)
+CORS_ORIGINS=["http://localhost:3000","http://localhost:5173"]
 
 # AI (optional, can be configured in the web UI)
 OPENAI_API_KEY=

backend/app/api/models.py

@@ -24,7 +24,6 @@ async def create_model(data: AIModelCreate, db: AsyncSession = Depends(get_db)):
     db.add(model)
     await db.flush()
     await db.refresh(model)
-    model.api_key = "***"
     return model
 
 
@@ -41,8 +40,6 @@ async def list_models(
     query = query.offset(skip).limit(limit).order_by(AIModel.created_at.desc())
     result = await db.execute(query)
     models = result.scalars().all()
-    for m in models:
-        m.api_key = "***"
     return models
 
 
@@ -52,7 +49,6 @@ async def get_model(model_id: str, db: AsyncSession = Depends(get_db)):
     model = result.scalar_one_or_none()
     if not model:
         raise HTTPException(status_code=404, detail="模型不存在")
-    model.api_key = "***"
     return model
 
 
@@ -72,7 +68,6 @@ async def update_model(model_id: str, data: AIModelUpdate, db: AsyncSession = De
 
     await db.flush()
     await db.refresh(model)
-    model.api_key = "***"
     return model
 
 
@@ -95,5 +90,4 @@ async def toggle_model(model_id: str, data: AIModelToggle, db: AsyncSession = De
     model.is_enabled = data.is_enabled
     await db.flush()
     await db.refresh(model)
-    model.api_key = "***"
     return model

backend/app/api/templates.py

@@ -1,4 +1,5 @@
 import uuid
+import bleach
 from fastapi import APIRouter, Depends, HTTPException, UploadFile, File, Form, Query
 from fastapi.responses import Response
 from sqlalchemy.ext.asyncio import AsyncSession
@@ -29,8 +30,6 @@ ALLOWED_ATTRS = {
     "table": ["style"],
 }
 
-import bleach
-
 router = APIRouter(prefix="/templates", tags=["模板管理"])
 
 

backend/app/schemas/ai_model.py

@@ -1,6 +1,6 @@
 import uuid
 from datetime import datetime
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, field_serializer
 
 
 class AIModelCreate(BaseModel):
@@ -32,7 +32,7 @@ class AIModelResponse(BaseModel):
     name: str
     provider: str
     endpoint: str
-    api_key: str = Field(default="***")
+    api_key: str
     extra_params: dict
     is_enabled: bool
     remark: str | None = None
@@ -40,3 +40,7 @@ class AIModelResponse(BaseModel):
     updated_at: datetime
 
     model_config = {"from_attributes": True}
+
+    @field_serializer("api_key")
+    def mask_api_key(self, value: str) -> str:
+        return "***"

backend/app/services/ai_adapter.py

@@ -105,12 +105,16 @@ class CustomAdapter(AIAdapter):
         if ref_content:
             user_content = f"参考以下内容：\n{ref_content}\n\n任务：{prompt}"
 
+        messages = [{"role": "system", "content": "你是一个专业的文档内容生成助手。"}]
+        messages.append({"role": "user", "content": user_content})
+
         body = {
-            "prompt": user_content,
+            "model": extra.get("model", "gpt-3.5-turbo"),
+            "messages": messages,
             "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")})
+        body.update({k: v for k, v in extra.items() if k not in ("model", "messages", "max_tokens", "temperature")})
 
         return {
             "url": model_config["endpoint"],
@@ -153,7 +157,12 @@ async def call_ai_model(model_config: dict, prompt: str, ref_content: str | None
 
     timeout = model_config.get("extra_params", {}).get("timeout", 120)
 
-    async with httpx.AsyncClient(timeout=timeout) as client:
+    try:
+        async with httpx.AsyncClient(
+            timeout=timeout,
+            proxy=None,
+            trust_env=False,
+        ) as client:
             response = await client.post(
                 request["url"],
                 headers=request["headers"],
@@ -161,3 +170,12 @@ async def call_ai_model(model_config: dict, prompt: str, ref_content: str | None
             )
             response.raise_for_status()
             return adapter.parse_response(response.json())
+    except httpx.HTTPStatusError as e:
+        detail = e.response.text[:500] if e.response else str(e)
+        raise RuntimeError(f"AI 服务返回错误 ({e.response.status_code}): {detail}")
+    except httpx.TimeoutException:
+        raise RuntimeError("AI 调用超时")
+    except httpx.ConnectError as e:
+        raise RuntimeError(f"无法连接 AI 服务: {e}")
+    except Exception as e:
+        raise RuntimeError(f"AI 调用异常: {str(e)}")