* Relative overlap = intersection area / min(area_A, area_B). * 1.0 means one element is fully covered by the other.

Per-tier number of "stuck-signal / extra-attempt units" required * before the literal answer may be released. Beginner releases fast, * advanced a

* Render an audio element.

* Replace a <video> with an <img> showing its poster (or, if no poster is set, * its current decoded frame drawn onto a canvas). Copies the video's i

Resolve blip opacity from alphaModFix / alphaMod / alphaOff modifiers.

* Resolve overall image opacity from OOXML blip alpha modifiers. * Same logic as ImageRenderer.resolveBlipOpacity: * - alphaModFix amt="N" * - alph

* Map our width x height to DashScope size format "WxH". * Common sizes: 1024*1024, 1280*720, 1664*928, 1120*1440, etc.

* Resolve fallback image from the mc:Fallback branch blipFill embed.

* Resolve image ID references in src field to actual base64 URLs * * AI generates: { type: "image", src: "img_1", ... } * This function replaces: {

* Resolve fill from layout or master placeholder when the slide shape has no explicit fill. * Checks if the shape's spPr explicitly declares <a:noFil

Match a browser language code (e.g. 'en', 'zh-TW') to a supported locale

* Look up a video/image element's generated media reference in the current stage's scenes. * Returns mediaRef first, then legacy src if it's a medi

* Resolve an embedded .docx package from the slide's rels + presentation embeddings. * Returns the word/document.xml content string, or null.

* Resolve image-level hyperlink from hlinkClick on cNvPr. * Mirrors link resolution logic in ShapeRenderer / ImageRenderer.

* Walk through all slide nodes (including group children recursively) * and fill in missing position/size from layout/master placeholders.

* Resolve clip geometry from spPr (prstGeom or custGeom). * Many decks use custGeom circles on p:pic instead of prst="ellipse".

* Map our aspect ratio + size to Seedream size format "WxH". * Seedream requires minimum 3,686,400 pixels total. * Common sizes: 2048x2048 (2K), 256

Resolve shape blipFill to embedded image data + opacity for JSON.

* Resolve the picture soft-edge feather radius (px) from * `spPr > effectLst > softEdge@rad`. PowerPoint feathers the image's alpha to * transparent

* Resolve target agent from @mention, or fallback to question agent for plain messages

* Compute CSS text-shadow from <a:effectLst><a:outerShdw> inside a rPr. PPT * cover titles routinely use this for the soft drop shadow under big chap

* Run generation for one agent. Streams agent_start, text_delta, * action, and agent_end events via config.writer().

Mark the current (last) text item as complete — no more appends expected.

* Use the official Brave Search API (requires API key). * Docs: https://api.search.brave.com/app/documentation/web-search

* Fallback: scrape the public Brave Search HTML page (no API key needed).

Seed IndexedDB with stage + 3 scenes using raw IndexedDB API

* Standard CCW segment-segment intersection (proper crossing only).