dashscope 1.24.1__py3-none-any.whl → 1.24.3__py3-none-any.whl

dashscope/__init__.py

@@ -7,7 +7,7 @@ from dashscope.aigc.code_generation import CodeGeneration
 from dashscope.aigc.conversation import Conversation, History, HistoryItem
 from dashscope.aigc.generation import AioGeneration, Generation
 from dashscope.aigc.image_synthesis import ImageSynthesis
-from dashscope.aigc.multimodal_conversation import MultiModalConversation
+from dashscope.aigc.multimodal_conversation import MultiModalConversation, AioMultiModalConversation
 from dashscope.aigc.video_synthesis import VideoSynthesis
 from dashscope.app.application import Application
 from dashscope.assistants import Assistant, AssistantList, Assistants
@@ -60,6 +60,7 @@ __all__ = [
     MultiModalEmbeddingItemText,
     SpeechSynthesizer,
     MultiModalConversation,
+    AioMultiModalConversation,
     BatchTextEmbedding,
     BatchTextEmbeddingResponse,
     Understanding,

dashscope/aigc/__init__.py

@@ -3,7 +3,7 @@
 from .conversation import Conversation, History, HistoryItem
 from .generation import Generation
 from .image_synthesis import ImageSynthesis
-from .multimodal_conversation import MultiModalConversation
+from .multimodal_conversation import MultiModalConversation, AioMultiModalConversation
 from .video_synthesis import VideoSynthesis
 
 __all__ = [
@@ -13,5 +13,6 @@ __all__ = [
     History,
     ImageSynthesis,
     MultiModalConversation,
+    AioMultiModalConversation,
     VideoSynthesis,
 ]

dashscope/aigc/image_synthesis.py

@@ -4,7 +4,7 @@ from typing import Any, Dict, List, Union
 
 from dashscope.api_entities.dashscope_response import (DashScopeAPIResponse,
                                                        ImageSynthesisResponse)
-from dashscope.client.base_api import BaseAsyncApi
+from dashscope.client.base_api import BaseAsyncApi, BaseApi, BaseAsyncAioApi, BaseAioApi
 from dashscope.common.constants import IMAGES, NEGATIVE_PROMPT, PROMPT
 from dashscope.common.error import InputRequired
 from dashscope.common.utils import _get_task_group_and_task
@@ -90,7 +90,7 @@ class ImageSynthesis(BaseAsyncApi):
                             **kwargs)
 
     @classmethod
-    def async_call(cls,
+    def sync_call(cls,
                    model: str,
                    prompt: Any,
                    negative_prompt: Any = None,
@@ -105,83 +105,73 @@ class ImageSynthesis(BaseAsyncApi):
                    mask_image_url: str = None,
                    base_image_url: str = None,
                    **kwargs) -> ImageSynthesisResponse:
-        """Create a image(s) synthesis task, and return task information.
-
-        Args:
-            model (str): The model, reference ``Models``.
-            prompt (Any): The prompt for image(s) synthesis.
-            negative_prompt (Any): The negative_prompt. Defaults to None.
-            images (List[str]): The input list of images url.
-            api_key (str, optional): The api api_key. Defaults to None.
-            sketch_image_url (str, optional): Only for wanx-sketch-to-image-v1.
-                Defaults to None.
-            workspace (str): The dashscope workspace id.
-            extra_input (Dict): The extra input parameters.
-            task (str): The task of api, ref doc.
-            function (str): The specific functions to be achieved. like:
-                colorization,super_resolution,expand,remove_watermaker,doodle,
-                description_edit_with_mask,description_edit,stylization_local,stylization_all
-            base_image_url (str): Enter the URL address of the target edited image.
-            mask_image_url (str): Provide the URL address of the image of the marked area by the user. It should be consistent with the image resolution of the base_image_url.
-            **kwargs(wanx-v1):
-                n(int, `optional`): Number of images to synthesis.
-                size: The output image(s) size, Default 1024*1024
-                similarity(float, `optional`): The similarity between the
-                    output image and the input image.
-                sketch_weight(int, optional): How much the input sketch
-                    affects the output image[0-10], only for wanx-sketch-to-image-v1. # noqa E501
-                    Default 10.
-                realisticness(int, optional): The realisticness of the output
-                    image[0-10], only for wanx-sketch-to-image-v1. Default 5
-
-        Raises:
-            InputRequired: The prompt cannot be empty.
-
-        Returns:
-            DashScopeAPIResponse: The image synthesis
-                task id in the response.
         """
+        Note: This method currently now only supports wan2.2-t2i-flash and wan2.2-t2i-plus.
+            Using other models will result in an error，More raw image models may be added for use later
+        """
+        task_group, f = _get_task_group_and_task(__name__)
+        inputs, kwargs, task = cls._get_input(model, prompt, negative_prompt,
+                                             images, api_key, sketch_image_url,
+                                             ref_img, extra_input, task, function,
+                                             mask_image_url, base_image_url, **kwargs)
+        response = BaseApi.call(model, inputs, task_group, task, f, api_key, workspace, **kwargs)
+        return ImageSynthesisResponse.from_api_response(response)
+
+    @classmethod
+    def _get_input(cls,
+                    model: str,
+                    prompt: Any,
+                    negative_prompt: Any = None,
+                    images: List[str] = None,
+                    api_key: str = None,
+                    sketch_image_url: str = None,
+                    ref_img: str = None,
+                    extra_input: Dict = None,
+                    task: str = None,
+                    function: str = None,
+                    mask_image_url: str = None,
+                    base_image_url: str = None,
+                    **kwargs):
         if prompt is None or not prompt:
             raise InputRequired('prompt is required!')
-        task_group, f = _get_task_group_and_task(__name__)
-        input = {PROMPT: prompt}
+        inputs = {PROMPT: prompt}
         has_upload = False
         if negative_prompt is not None:
-            input[NEGATIVE_PROMPT] = negative_prompt
+            inputs[NEGATIVE_PROMPT] = negative_prompt
         if images is not None:
-            input[IMAGES] = images
+            inputs[IMAGES] = images
         if sketch_image_url is not None and sketch_image_url:
             is_upload, sketch_image_url = check_and_upload_local(
                 model, sketch_image_url, api_key)
             if is_upload:
                 has_upload = True
-            input['sketch_image_url'] = sketch_image_url
+            inputs['sketch_image_url'] = sketch_image_url
         if ref_img is not None and ref_img:
             is_upload, ref_img = check_and_upload_local(
                 model, ref_img, api_key)
             if is_upload:
                 has_upload = True
-            input['ref_img'] = ref_img
+            inputs['ref_img'] = ref_img
 
         if function is not None and function:
-            input['function'] = function
+            inputs['function'] = function
 
         if mask_image_url is not None and mask_image_url:
             is_upload, res_mask_image_url = check_and_upload_local(
                 model, mask_image_url, api_key)
             if is_upload:
                 has_upload = True
-            input['mask_image_url'] = res_mask_image_url
+            inputs['mask_image_url'] = res_mask_image_url
 
         if base_image_url is not None and base_image_url:
             is_upload, res_base_image_url = check_and_upload_local(
                 model, base_image_url, api_key)
             if is_upload:
                 has_upload = True
-            input['base_image_url'] = res_base_image_url
+            inputs['base_image_url'] = res_base_image_url
 
         if extra_input is not None and extra_input:
-            input = {**input, **extra_input}
+            inputs = {**inputs, **extra_input}
 
         if has_upload:
             headers = kwargs.pop('headers', {})
@@ -193,13 +183,72 @@ class ImageSynthesis(BaseAsyncApi):
         if model is not None and model and 'imageedit' in model:
             task = 'image2image'
 
+        return inputs, kwargs, task
+
+    @classmethod
+    def async_call(cls,
+                   model: str,
+                   prompt: Any,
+                   negative_prompt: Any = None,
+                   images: List[str] = None,
+                   api_key: str = None,
+                   sketch_image_url: str = None,
+                   ref_img: str = None,
+                   workspace: str = None,
+                   extra_input: Dict = None,
+                   task: str = None,
+                   function: str = None,
+                   mask_image_url: str = None,
+                   base_image_url: str = None,
+                   **kwargs) -> ImageSynthesisResponse:
+        """Create a image(s) synthesis task, and return task information.
+
+        Args:
+            model (str): The model, reference ``Models``.
+            prompt (Any): The prompt for image(s) synthesis.
+            negative_prompt (Any): The negative_prompt. Defaults to None.
+            images (List[str]): The input list of images url.
+            api_key (str, optional): The api api_key. Defaults to None.
+            sketch_image_url (str, optional): Only for wanx-sketch-to-image-v1.
+                Defaults to None.
+            workspace (str): The dashscope workspace id.
+            extra_input (Dict): The extra input parameters.
+            task (str): The task of api, ref doc.
+            function (str): The specific functions to be achieved. like:
+                colorization,super_resolution,expand,remove_watermaker,doodle,
+                description_edit_with_mask,description_edit,stylization_local,stylization_all
+            base_image_url (str): Enter the URL address of the target edited image.
+            mask_image_url (str): Provide the URL address of the image of the marked area by the user. It should be consistent with the image resolution of the base_image_url.
+            **kwargs(wanx-v1):
+                n(int, `optional`): Number of images to synthesis.
+                size: The output image(s) size, Default 1024*1024
+                similarity(float, `optional`): The similarity between the
+                    output image and the input image.
+                sketch_weight(int, optional): How much the input sketch
+                    affects the output image[0-10], only for wanx-sketch-to-image-v1. # noqa E501
+                    Default 10.
+                realisticness(int, optional): The realisticness of the output
+                    image[0-10], only for wanx-sketch-to-image-v1. Default 5
+
+        Raises:
+            InputRequired: The prompt cannot be empty.
+
+        Returns:
+            DashScopeAPIResponse: The image synthesis
+                task id in the response.
+        """
+        task_group, f = _get_task_group_and_task(__name__)
+        inputs, kwargs, task = cls._get_input(model, prompt, negative_prompt,
+                                             images, api_key, sketch_image_url,
+                                             ref_img, extra_input, task, function,
+                                             mask_image_url, base_image_url, **kwargs)
         response = super().async_call(
             model=model,
             task_group=task_group,
             task=task,
             function=f,
             api_key=api_key,
-            input=input,
+            input=inputs,
             workspace=workspace,
             **kwargs)
         return ImageSynthesisResponse.from_api_response(response)
@@ -306,3 +355,259 @@ class ImageSynthesis(BaseAsyncApi):
                             api_key=api_key,
                             workspace=workspace,
                             **kwargs)
+
+class AioImageSynthesis(BaseAsyncAioApi):
+    @classmethod
+    async def call(cls,
+                  model: str,
+                  prompt: Any,
+                  negative_prompt: Any = None,
+                  images: List[str] = None,
+                  api_key: str = None,
+                  sketch_image_url: str = None,
+                  ref_img: str = None,
+                  workspace: str = None,
+                  extra_input: Dict = None,
+                  task: str = None,
+                  function: str = None,
+                  mask_image_url: str = None,
+                  base_image_url: str = None,
+                  **kwargs) -> ImageSynthesisResponse:
+        """Call image(s) synthesis service and get result.
+
+        Args:
+            model (str): The model, reference ``Models``.
+            prompt (Any): The prompt for image(s) synthesis.
+            negative_prompt (Any): The negative_prompt. Defaults to None.
+            images (List[str]): The input list of images url,
+                currently not supported.
+            api_key (str, optional): The api api_key. Defaults to None.
+            sketch_image_url (str, optional): Only for wanx-sketch-to-image-v1,
+                can be local file.
+                Defaults to None.
+            workspace (str): The dashscope workspace id.
+            extra_input (Dict): The extra input parameters.
+            task (str): The task of api, ref doc.
+            function (str): The specific functions to be achieved. like:
+                colorization,super_resolution,expand,remove_watermaker,doodle,
+                description_edit_with_mask,description_edit,stylization_local,stylization_all
+            base_image_url (str): Enter the URL address of the target edited image.
+            mask_image_url (str): Provide the URL address of the image of the marked area by the user. It should be consistent with the image resolution of the base_image_url.
+            **kwargs:
+                n(int, `optional`): Number of images to synthesis.
+                size(str, `optional`): The output image(s) size(width*height).
+                similarity(float, `optional`): The similarity between the
+                    output image and the input image
+                sketch_weight(int, optional): How much the input sketch
+                    affects the output image[0-10], only for wanx-sketch-to-image-v1. # noqa E501
+                    Default 10.
+                realisticness(int, optional): The realisticness of the output
+                    image[0-10], only for wanx-sketch-to-image-v1. Default 5
+
+        Raises:
+            InputRequired: The prompt cannot be empty.
+
+        Returns:
+            ImageSynthesisResponse: The image(s) synthesis result.
+        """
+        task_group, f = _get_task_group_and_task(__name__)
+        inputs, kwargs, task = ImageSynthesis._get_input(model, prompt, negative_prompt,
+                                               images, api_key, sketch_image_url,
+                                               ref_img, extra_input, task, function,
+                                               mask_image_url, base_image_url, **kwargs)
+        response = await super().call(model, inputs, task_group, task, f, api_key, workspace, **kwargs)
+        return ImageSynthesisResponse.from_api_response(response)
+
+    @classmethod
+    async def sync_call(cls,
+                   model: str,
+                   prompt: Any,
+                   negative_prompt: Any = None,
+                   images: List[str] = None,
+                   api_key: str = None,
+                   sketch_image_url: str = None,
+                   ref_img: str = None,
+                   workspace: str = None,
+                   extra_input: Dict = None,
+                   task: str = None,
+                   function: str = None,
+                   mask_image_url: str = None,
+                   base_image_url: str = None,
+                   **kwargs) -> ImageSynthesisResponse:
+        """
+        Note: This method currently now only supports wan2.2-t2i-flash and wan2.2-t2i-plus.
+            Using other models will result in an error，More raw image models may be added for use later
+        """
+        task_group, f = _get_task_group_and_task(__name__)
+        inputs, kwargs, task = ImageSynthesis._get_input(model, prompt, negative_prompt,
+                                             images, api_key, sketch_image_url,
+                                             ref_img, extra_input, task, function,
+                                             mask_image_url, base_image_url, **kwargs)
+        response = await BaseAioApi.call(model, inputs, task_group, task, f, api_key, workspace, **kwargs)
+        return ImageSynthesisResponse.from_api_response(response)
+
+    @classmethod
+    async def async_call(cls,
+                  model: str,
+                  prompt: Any,
+                  negative_prompt: Any = None,
+                  images: List[str] = None,
+                  api_key: str = None,
+                  sketch_image_url: str = None,
+                  ref_img: str = None,
+                  workspace: str = None,
+                  extra_input: Dict = None,
+                  task: str = None,
+                  function: str = None,
+                  mask_image_url: str = None,
+                  base_image_url: str = None,
+                  **kwargs) -> ImageSynthesisResponse:
+        """Create a image(s) synthesis task, and return task information.
+
+        Args:
+            model (str): The model, reference ``Models``.
+            prompt (Any): The prompt for image(s) synthesis.
+            negative_prompt (Any): The negative_prompt. Defaults to None.
+            images (List[str]): The input list of images url.
+            api_key (str, optional): The api api_key. Defaults to None.
+            sketch_image_url (str, optional): Only for wanx-sketch-to-image-v1.
+                Defaults to None.
+            workspace (str): The dashscope workspace id.
+            extra_input (Dict): The extra input parameters.
+            task (str): The task of api, ref doc.
+            function (str): The specific functions to be achieved. like:
+                colorization,super_resolution,expand,remove_watermaker,doodle,
+                description_edit_with_mask,description_edit,stylization_local,stylization_all
+            base_image_url (str): Enter the URL address of the target edited image.
+            mask_image_url (str): Provide the URL address of the image of the marked area by the user. It should be consistent with the image resolution of the base_image_url.
+            **kwargs(wanx-v1):
+                n(int, `optional`): Number of images to synthesis.
+                size: The output image(s) size, Default 1024*1024
+                similarity(float, `optional`): The similarity between the
+                    output image and the input image.
+                sketch_weight(int, optional): How much the input sketch
+                    affects the output image[0-10], only for wanx-sketch-to-image-v1. # noqa E501
+                    Default 10.
+                realisticness(int, optional): The realisticness of the output
+                    image[0-10], only for wanx-sketch-to-image-v1. Default 5
+
+        Raises:
+            InputRequired: The prompt cannot be empty.
+
+        Returns:
+            DashScopeAPIResponse: The image synthesis
+                task id in the response.
+        """
+        task_group, f = _get_task_group_and_task(__name__)
+        inputs, kwargs, task = ImageSynthesis._get_input(model, prompt, negative_prompt,
+                                               images, api_key, sketch_image_url,
+                                               ref_img, extra_input, task, function,
+                                               mask_image_url, base_image_url, **kwargs)
+        response = await super().async_call(model, inputs, task_group, task, f, api_key, workspace, **kwargs)
+        return ImageSynthesisResponse.from_api_response(response)
+
+    @classmethod
+    async def fetch(cls,
+              task: Union[str, ImageSynthesisResponse],
+              api_key: str = None,
+              workspace: str = None,
+              **kwargs,) -> ImageSynthesisResponse:
+        """Fetch image(s) synthesis task status or result.
+
+        Args:
+            task (Union[str, ImageSynthesisResponse]): The task_id or
+                ImageSynthesisResponse return by async_call().
+            api_key (str, optional): The api api_key. Defaults to None.
+            workspace (str): The dashscope workspace id.
+
+        Returns:
+            ImageSynthesisResponse: The task status or result.
+        """
+        response = await super().fetch(task, api_key=api_key, workspace=workspace)
+        return ImageSynthesisResponse.from_api_response(response)
+
+    @classmethod
+    async def wait(cls,
+             task: Union[str, ImageSynthesisResponse],
+             api_key: str = None,
+             workspace: str = None,
+             **kwargs) -> ImageSynthesisResponse:
+        """Wait for image(s) synthesis task to complete, and return the result.
+
+        Args:
+            task (Union[str, ImageSynthesisResponse]): The task_id or
+                ImageSynthesisResponse return by async_call().
+            api_key (str, optional): The api api_key. Defaults to None.
+            workspace (str): The dashscope workspace id.
+
+        Returns:
+            ImageSynthesisResponse: The task result.
+        """
+        response = await super().wait(task, api_key, workspace=workspace)
+        return ImageSynthesisResponse.from_api_response(response)
+
+    @classmethod
+    async def cancel(cls,
+                task: Union[str, ImageSynthesisResponse],
+                api_key: str = None,
+                workspace: str = None,
+                **kwargs,) -> DashScopeAPIResponse:
+        """Cancel image synthesis task.
+        Only tasks whose status is PENDING can be canceled.
+
+        Args:
+            task (Union[str, ImageSynthesisResponse]): The task_id or
+                ImageSynthesisResponse return by async_call().
+            api_key (str, optional): The api api_key. Defaults to None.
+            workspace (str): The dashscope workspace id.
+
+        Returns:
+            DashScopeAPIResponse: The response data.
+        """
+        return await super().cancel(task, api_key, workspace=workspace)
+
+    @classmethod
+    async def list(cls,
+             start_time: str = None,
+             end_time: str = None,
+             model_name: str = None,
+             api_key_id: str = None,
+             region: str = None,
+             status: str = None,
+             page_no: int = 1,
+             page_size: int = 10,
+             api_key: str = None,
+             workspace: str = None,
+             **kwargs) -> DashScopeAPIResponse:
+        """List async tasks.
+
+        Args:
+            start_time (str, optional): The tasks start time,
+                for example: 20230420000000. Defaults to None.
+            end_time (str, optional): The tasks end time,
+                for example: 20230420000000. Defaults to None.
+            model_name (str, optional): The tasks model name. Defaults to None.
+            api_key_id (str, optional): The tasks api-key-id. Defaults to None.
+            region (str, optional): The service region,
+                for example: cn-beijing. Defaults to None.
+            status (str, optional): The status of tasks[PENDING,
+                RUNNING, SUCCEEDED, FAILED, CANCELED]. Defaults to None.
+            page_no (int, optional): The page number. Defaults to 1.
+            page_size (int, optional): The page size. Defaults to 10.
+            api_key (str, optional): The user api-key. Defaults to None.
+            workspace (str): The dashscope workspace id.
+
+        Returns:
+            DashScopeAPIResponse: The response data.
+        """
+        return await super().list(start_time=start_time,
+                            end_time=end_time,
+                            model_name=model_name,
+                            api_key_id=api_key_id,
+                            region=region,
+                            status=status,
+                            page_no=page_no,
+                            page_size=page_size,
+                            api_key=api_key,
+                            workspace=workspace,
+                            **kwargs)

dashscope/aigc/multimodal_conversation.py

@@ -5,7 +5,7 @@ from typing import Generator, List, Union
 
 from dashscope.api_entities.dashscope_response import \
     MultiModalConversationResponse
-from dashscope.client.base_api import BaseApi
+from dashscope.client.base_api import BaseAioApi, BaseApi
 from dashscope.common.error import InputRequired, ModelRequired
 from dashscope.common.utils import _get_task_group_and_task
 from dashscope.utils.oss_utils import preprocess_message_element
@@ -130,3 +130,124 @@ class MultiModalConversation(BaseApi):
                     if is_upload and not has_upload:
                         has_upload = True
         return has_upload
+
+
+class AioMultiModalConversation(BaseAioApi):
+    """Async MultiModal conversational robot interface.
+    """
+    task = 'multimodal-generation'
+    function = 'generation'
+
+    class Models:
+        qwen_vl_chat_v1 = 'qwen-vl-chat-v1'
+
+    @classmethod
+    async def call(
+        cls,
+        model: str,
+        messages: List,
+        api_key: str = None,
+        workspace: str = None,
+        **kwargs
+    ) -> Union[MultiModalConversationResponse, Generator[
+            MultiModalConversationResponse, None, None]]:
+        """Call the conversation model service asynchronously.
+
+        Args:
+            model (str): The requested model, such as 'qwen-multimodal-v1'
+            messages (list): The generation messages.
+                examples:
+                    [
+                        {
+                            "role": "system",
+                            "content": [
+                                {"text": "你是达摩院的生活助手机器人。"}
+                            ]
+                        },
+                        {
+                            "role": "user",
+                            "content": [
+                                {"image": "http://XXXX"},
+                                {"text": "这个图片是哪里？"},
+                            ]
+                        }
+                    ]
+            api_key (str, optional): The api api_key, can be None,
+                if None, will retrieve by rule [1].
+                [1]: https://help.aliyun.com/zh/dashscope/developer-reference/api-key-settings. # noqa E501
+            workspace (str): The dashscope workspace id.
+            **kwargs:
+                stream(bool, `optional`): Enable server-sent events
+                    (ref: https://developer.mozilla.org/en-US/docs/Web/API/Server-sent_events/Using_server-sent_events)  # noqa E501
+                    the result will back partially[qwen-turbo,bailian-v1].
+                max_length(int, `optional`): The maximum length of tokens to
+                    generate. The token count of your prompt plus max_length
+                    cannot exceed the model's context length. Most models
+                    have a context length of 2000 tokens[qwen-turbo,bailian-v1]. # noqa E501
+                top_p(float, `optional`): A sampling strategy, called nucleus
+                    sampling, where the model considers the results of the
+                    tokens with top_p probability mass. So 0.1 means only
+                    the tokens comprising the top 10% probability mass are
+                    considered[qwen-turbo,bailian-v1].
+                top_k(float, `optional`):
+
+        Raises:
+            InvalidInput: The history and auto_history are mutually exclusive.
+
+        Returns:
+            Union[MultiModalConversationResponse,
+                  Generator[MultiModalConversationResponse, None, None]]: If
+            stream is True, return Generator, otherwise MultiModalConversationResponse.
+        """
+        if (messages is None or not messages):
+            raise InputRequired('prompt or messages is required!')
+        if model is None or not model:
+            raise ModelRequired('Model is required!')
+        task_group, _ = _get_task_group_and_task(__name__)
+        msg_copy = copy.deepcopy(messages)
+        has_upload = cls._preprocess_messages(model, msg_copy, api_key)
+        if has_upload:
+            headers = kwargs.pop('headers', {})
+            headers['X-DashScope-OssResourceResolve'] = 'enable'
+            kwargs['headers'] = headers
+        input = {'messages': msg_copy}
+        response = await super().call(model=model,
+                                      task_group=task_group,
+                                      task=AioMultiModalConversation.task,
+                                      function=AioMultiModalConversation.function,
+                                      api_key=api_key,
+                                      input=input,
+                                      workspace=workspace,
+                                      **kwargs)
+        is_stream = kwargs.get('stream', False)
+        if is_stream:
+            return (MultiModalConversationResponse.from_api_response(rsp)
+                    async for rsp in response)
+        else:
+            return MultiModalConversationResponse.from_api_response(response)
+
+    @classmethod
+    def _preprocess_messages(cls, model: str, messages: List[dict],
+                             api_key: str):
+        """
+            messages = [
+                {
+                    "role": "user",
+                    "content": [
+                        {"image": ""},
+                        {"text": ""},
+                    ]
+                }
+            ]
+        """
+        has_upload = False
+        for message in messages:
+            content = message['content']
+            for elem in content:
+                if not isinstance(elem,
+                                  (int, float, bool, str, bytes, bytearray)):
+                    is_upload = preprocess_message_element(
+                        model, elem, api_key)
+                    if is_upload and not has_upload:
+                        has_upload = True
+        return has_upload