only for Image to Text

app.py

@@ -18,6 +18,7 @@ from fastapi.responses import JSONResponse
 from PIL import Image, UnidentifiedImageError
 from pymongo import MongoClient
 from pymongo.errors import PyMongoError, ServerSelectionTimeoutError
+from starlette.datastructures import UploadFile as StarletteUploadFile
 from transformers import (
     AutoModelForImageTextToText,
     AutoModelForSeq2SeqLM,
@@ -362,16 +363,16 @@ async def generate_caption(request: Request):
     except Exception as exc:
         raise AppError("Invalid request payload.", 422) from exc
 
-    uploads: list[UploadFile] = []
+    uploads: list[UploadFile | StarletteUploadFile] = []
     for key in ("files", "files[]", "file"):
         for value in form.getlist(key):
-            if isinstance(value, UploadFile):
+            if isinstance(value, (UploadFile, StarletteUploadFile)):
                 uploads.append(value)
 
     # Fallback for clients that send non-standard multipart keys.
     if not uploads:
         for _, value in form.multi_items():
-            if isinstance(value, UploadFile):
+            if isinstance(value, (UploadFile, StarletteUploadFile)):
                 uploads.append(value)
 
     if not uploads: