@Alpee Jain The current version of the custom vision or computer vision API's allows you to pass an image URL and get the tags or features associated with the image so that the response can be processed by the client. The API itself cannot directly analyze and provide the list of image URLs that are similar from a storage location. In an ideal scenario you would like to tag the images with different tags during training and then use the predict API to get the predicted tag and classify the image under that category or tag.
If you would like to still use one tag and check for similar images in your database, you can make individual calls of an image URL to computer vision analyze image API and process the responses from all image URLs and store the results in a DB so your product can give suggestions to your customers later. I hope this helps.