refactor(chat-models)!: Make MIME Type mandatory for base64 images in prompt

[davidmigloz]

To be compatible with other multimodal LLMs, the ChatMessageContentImage class now requires:

• data: Either a base64 encoded image data or a URL (for models that support it)
• mimeType: The IANA standard MIME type of the source data (e.g. image/png)

This is a breaking change if you were using GPT-4V.

Migration

With Base64 image

Rename url to data and specify mimeType.

Before:

final prompt = PromptValue.chat([
  ChatMessage.system(
    'You are a helpful assistant.',
  ),
  ChatMessage.human(
    ChatMessageContent.multiModal([
      ChatMessageContent.text('What fruit is this?'),
      ChatMessageContent.image(
        url: 'data:image/jpeg;base64,/9j/4AAQS...nSbsr69/Rdu1j//2Q==',
      ),
    ]),
  ),
]);

Now:

final prompt = PromptValue.chat([
  ChatMessage.system(
    'You are a helpful assistant.',
  ),
  ChatMessage.human(
    ChatMessageContent.multiModal([
      ChatMessageContent.text('What fruit is this?'),
      ChatMessageContent.image(
        mimeType: 'image/jpeg',
        data: '/9j/4AAQS...nSbsr69/Rdu1j//2Q==',
      ),
    ]),
  ),
]);

With URL

Just rename url to data.

Before:

ChatMessageContent.image(
  url: 'https://upload.wikimedia.org/wikipedia/commons/9/92/95apple.jpeg',
), 

Now:

ChatMessageContent.image(
  data: 'https://upload.wikimedia.org/wikipedia/commons/9/92/95apple.jpeg',
),