Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.clearmaas.com/llms.txt

Use this file to discover all available pages before exploring further.

Send an image as a content part with type: "image_url". Pass a publicly-reachable https:// URL — this is the most universally supported form. data:image/...;base64,... inline URIs work for OpenAI and Gemini targets. For Anthropic and xAI Grok models, prefer hosting the image behind an https:// URL or use the provider’s native HTTP shape (see Native Formats) when sending base64. 
resp = client.chat.completions.create(
    model="openai/gpt-4o-mini",
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "What's in this image?"},
            {"type": "image_url", "image_url": {"url": "https://example.com/cat.jpg"}},
        ],
    }],
)

​
Vision-capable model families

Vision works against any upstream model that accepts image input. Examples:
  • OpenAI gpt-4o* and gpt-4.1* family
  • Anthropic Claude 4 family (all current models)
  • Google Gemini multimodal (gemini-{2.5,3,3.1}-{flash,pro})
  • xAI Grok 4 family (vision is built into general Grok 4 chat models)

​
Size limits

Each upstream provider enforces its own per-image size cap (typically in the single-digit-MB range for inline base64 and higher for hosted URLs). ClearMaas respects the upstream’s limit, so exceeding it surfaces as a 400 from the provider. Check the upstream’s current vision documentation for the exact number.