Multimodal

PDFs

PDF/document input isn't handled today. A type:"file" content part isn't recognized or gated by capability — it's forwarded as-is to whatever provider your request routes to, so behavior depends entirely on whether that provider's own API happens to accept the shape. Don't build against it.

This also means the OpenRouter-style top-level plugins array (used there to select a PDF parser engine, e.g. mistral-ocr) is silently dropped before the request reaches a provider — sending it doesn't error, it just does nothing.

What to do today

Convert pages to images and send them through image understanding, or extract the text yourself (e.g. with a PDF-parsing library) and inline it as a plain text content part.