Extract video bytes from content parts, rewriting to the template's
{type:"video"} form. Accepts {type:"video_url", video_url:{url}} (data:
or http(s), same SSRF guard as images with the larger video body cap)
and {type:"video"} with base64 data. Decode to frames happens later
(packages/inference/src/input/vision/video-frames.ts — the AVFoundation sidecar).
Extract video bytes from content parts, rewriting to the template's {type:"video"} form. Accepts {type:"video_url", video_url:{url}} (data: or http(s), same SSRF guard as images with the larger video body cap) and {type:"video"} with base64
data. Decode to frames happens later (packages/inference/src/input/vision/video-frames.ts — the AVFoundation sidecar).