fix(ai): allow multimodal message content to fix image upload 422

- AIMessage.content: Union[str, List[dict]] for text-only and multimodal (e.g. image_url)
- Image requests with content array now pass validation and are forwarded to AI

app/schemas.py

@@ -1,5 +1,5 @@
 from pydantic import BaseModel, ConfigDict
-from typing import List, Optional
+from typing import List, Optional, Union
 from datetime import datetime
 
 # Heir Schemas
@@ -80,10 +80,10 @@ class AssetDelete(BaseModel):
 class DeclareGuale(BaseModel):
     username: str
 
-# AI Proxy Schemas
+# AI Proxy Schemas (content: str for text-only, list for multimodal e.g. image_url)
 class AIMessage(BaseModel):
     role: str
-    content: str
+    content: Union[str, List[dict]]
 
 class AIRequest(BaseModel):
     messages: List[AIMessage]