Azure cv

AzureCVToolSpec

基础类：BaseToolSpec

Azure 认知视觉工具规范。

源码位于 llama-index-integrations/tools/llama-index-tools-azure-cv/llama_index/tools/azure_cv/base.py

class AzureCVToolSpec(BaseToolSpec):
    """Azure Cognitive Vision tool spec."""

    spec_functions = ["process_image"]

    def __init__(
        self,
        resource: str,
        api_key: str,
        language: Optional[str] = "en",
        api_version: Optional[str] = "2023-04-01-preview",
    ) -> None:
        """Initialize with parameters."""
        self.api_key = api_key
        self.cv_url = CV_URL_TMPL.format(resource=resource)
        self.language = language
        self.api_version = api_version

    def process_image(self, url: str, features: List[str]):
        """
        This tool accepts an image url or file and can process and return a variety of text depending on the use case.
        You can use the features argument to configure what text you want returned.

        Args:
            url (str): The url for the image to caption
            features (List[str]): Instructions on how to process the image. Valid keys are tags, objects, read, caption

        """
        response = requests.post(
            f'{self.cv_url}?features={",".join(features)}&language={self.language}&api-version={self.api_version}',
            headers={"Ocp-Apim-Subscription-Key": self.api_key},
            json={"url": url},
        )
        response_json = response.json()
        if "read" in features:
            response_json["readResult"] = response_json["readResult"]["content"]

        return response_json

process_image

process_image(url: str, features: List[str])

此工具接受图像 URL 或文件，并可根据用例处理并返回各种文本。您可以使用 features 参数配置您想要返回的文本。

参数

名称	类型	描述	默认值
url	str	要生成字幕的图像 URL	必需
features	列表[字符串]	关于如何处理图像的说明。有效键包括 tags, objects, read, caption	必需

源码位于 llama-index-integrations/tools/llama-index-tools-azure-cv/llama_index/tools/azure_cv/base.py

def process_image(self, url: str, features: List[str]):
    """
    This tool accepts an image url or file and can process and return a variety of text depending on the use case.
    You can use the features argument to configure what text you want returned.

    Args:
        url (str): The url for the image to caption
        features (List[str]): Instructions on how to process the image. Valid keys are tags, objects, read, caption

    """
    response = requests.post(
        f'{self.cv_url}?features={",".join(features)}&language={self.language}&api-version={self.api_version}',
        headers={"Ocp-Apim-Subscription-Key": self.api_key},
        json={"url": url},
    )
    response_json = response.json()
    if "read" in features:
        response_json["readResult"] = response_json["readResult"]["content"]

    return response_json