feat(provider/perplexity): add support for return_images

[kevinwolfcr]

Adds support for return_images to the Perplexity Provider

Note

As per @shaper's recommendation, I'm temporarily adding the logic to extract and pass the options to the underlying request in openai-compatible-chat-language-model.ts, since the Perplexity provider extends it.

[shaper]

Hi there, thank you for the contribution! I just sent #4618 which should allow us to do the input part of this PR as:

  const result = await generateText({
    model: perplexity('sonar-pro'),
    prompt: 'Invent a new holiday and describe its traditions.',
    providerOptions: {
      perplexity: {
        request_images: true,
      },
    },
  });

and similar for streamText. This should mean no changes needed for openai-compatible. I see you've kindly also done the extractor logic and tests, so I will work with you to get this landed once the other change is in (likely tomorrow).

Goal is to get you operational ASAP!

[kevinwolfcr]

@shaper changes done. I removed the returnImages option from the openai-compatible-chat-language-model.ts and added it to the providerOptions object in the perplexity-with-images example for streamText and generateText.

[shaper]

@shaper changes done. I removed the returnImages option from the openai-compatible-chat-language-model.ts and added it to the providerOptions object in the perplexity-with-images example for streamText and generateText.

Thanks, I am looking at this

[shaper]

Continuing this work (inclusive of your commits to track credit) in #4651