Skip to main content
Anannas supports multiple input modalities beyond text, so you can send images, PDFs, and audio files to compatible models through the same /chat/completions endpoint. This enables rich, multimodal interactions with minimal integration overhead.

​
Supported Modalities

  • Images - Send images to vision-enabled models for tasks like analysis, captioning, OCR, and visual Q&A. Supports: URL-based images, base64-encoded images
    Read More about Image Inputs ->
  • PDFs - Process PDF documents seamlessly. Anannas extracts text and handles both text-based PDFs and scanned files. Accepted formats: URL or base64-encoded.
    Read More about PDF processing ->
  • Audio (Not Yet Supported) - You can technically send audio in requests, but Anthropic models do not process it. No error will be thrown, it will simply be ignored. We recommend avoiding audio inputs for now until support is expanded.
    Read More about Audio Input ->

​
Getting Started

All multimodal inputs use /chat/completions with the messages array.
Specify the content type for each input:
  1. Images -> image_url
  2. PDFs -> file
  3. Audio -> audio
You can combine multiple inputs in a single request. The maximum number of files depends on the model/provider.

​
Model Compatibility

  1. Images -> Supported (URL + base64)
  2. PDFs -> Supported (URL + base64)
  3. Audio -> Ignored (no error, no output)
You can combine multiple modalities in a single request, and the number of files you can send varies by provider and model.

​
Input Format Support

URLs (Recommended for public content)
  1. Images: https://example.com/image.jpg
  2. PDFs: https://example.com/document.pdf
  3. Audio: ❌ Ignored
Base64 Encoding (Required for local/private files)
  1. Images: data:image/jpeg;base64,{base64_data}
  2. PDFs: data:application/pdf;base64,{base64_data}
  3. Audio: Ignored
URLs are preferred for large files since they avoid payload bloat.Use base64 encoding when working with local or private files.

​
FAQs

Yes. You can mix text, images, and PDFs in the same request. The model will process them together.
Audio inputs are accepted but not processed. You won’t see an error, but the model won’t respond to the audio content.
Yes, audio support is on the roadmap. Stay tuned for updates.
Was this page helpful?