sonnedu/ai-model-icons

AI Model Icons is a free, ready to use library of brand and product icons for AI companies, large language models, assistants, and platforms. The README is mainly in Chinese with some English headings. At the time of writing the project lists 148 icon entries spanning model makers, products, assistants, platforms, and local runtimes, with 514 aliases that let you look an icon up by short name, lowercase string, Chinese name, model name, or product name. About 15 entries are still placeholder icons waiting for a clean source. For each icon the repo stores both an SVG and a set of PNG rasters at common sizes: 16, 32, 48, 180, 192, 512, and 1024 pixels. There are also Apple imagesets, Android adaptive icon files, and a web manifest layout, which means the same icon can be dropped into an iOS app, an Android app, or a web favicon setup without further conversion. Each catalog entry also gets quality and confidence labels so a consumer can tell whether the icon is a clean vector or a community guess. The README shows two ways to use the data. The static way is suited to GitHub Pages, CDNs, or any front end without a server: you fetch catalog/models.json plus catalog/aliases.json, normalize the user's input to lowercase with punctuation stripped, look up the id in the aliases map, and read the icon path from the catalog. A short JavaScript snippet shows the whole flow. The dynamic way runs the bundled Node API on port 8787 by default, with endpoints like /api/resolve?q=grok and /api/assets?q=grok that return JSON with id, owner, icon paths, and per platform asset profiles. The README is upfront about a few rough edges. Android VectorDrawable files are not arbitrary SVGs, so the API ships profile XML and PNG caches and points users at Android Studio's Vector Asset tool for complex vectors. There is also a table of matching examples to clarify how queries like z.ai, copilot, github copilot, and doubao are mapped to the right brand and parent. Maintenance is script driven. npm run build:aliases regenerates the alias index, npm run build:raster rebuilds the PNGs, npm run check runs the failure only validation, npm run audit produces a semantic report on missing icons, and npm run serve starts the API. Lower level node scripts can rebuild the catalog, sync brand icons from upstream sources, and resolve a single query from the command line.