madeye/bangumi-downloader

Bangumi Torrent Finder is a web application that searches multiple anime torrent sources at once and shows results in a single unified list. Instead of checking bangumi.moe, acg.rip, dmhy, and nyaa.si separately, you type one search query and the app queries all four sources in parallel, removes duplicates, and groups them by series. The app supports batch operations on search results: copy multiple magnet links at once, open torrents directly, or download all selected .torrent files bundled as a single zip archive. A two stage response model shows initial results immediately and then quietly refines the grouping in the background, so the page does not block while waiting. An optional LLM integration improves how results are grouped by series, and the README notes compatibility with Anthropic protocol endpoints such as Kimi and MiniMax. Without the LLM configured, the app falls back to title based heuristics. Search results are cached in SQLite with a configurable time to live, so repeated queries for the same series are served instantly. All requests to the upstream torrent sources are made server side, keeping provider API tokens out of the browser. The app is built with Next.js using the App Router, along with React and TypeScript. You set it up with npm install and npm run dev.