Kedou Media Workflow

Core Rule

Use Kedou only for content where the user has rights, permission, membership access, or a clear personal learning/research basis. Never request, store, or echo raw cookies, auth headers, account tokens, paid-content URLs, or private media links.

Decision Tree

1. Single public video or image: use Kedou web parser.
   • Copy browser URL or app share link.
   • Paste into Kedou.
   • Choose direct parse for the current visible media.
   • Choose list parse when the page contains a playlist, collection, homepage, or media list.
2. Batch videos, homepage, playlist, favorites, search, or high-resolution work: use Kedou desktop downloader.
   • Prefer the downloader for repeated jobs, batch input, playlist/homepage extraction, subtitles/danmaku in bulk, and sniffing.
   • Batch input expects one URL per line.
   • Keep task count small, usually 2 or 4.
3. Subtitles, danmaku, lyrics, or text only: use the matching Kedou extraction page first.
   • Subtitle pages return language-specific files, usually .srt, with view/download options.
   • Danmaku pages return platform danmaku where supported.
   • Text extraction pages return article/copy/image material, often Markdown-like.
4. Need fresh docs or current platform support: use $opencli-browser-first to read Kedou public pages, then update the workflow. Do not rely on stale memory for platform-specific support changes.

Downloader Operating Notes

• Use clipboard import for one copied link.
• Use manual link input when a URL or request headers must be supplied locally.
• Use batch input for multiple URLs, one URL per line.
• Use homepage/list workflows for supported user spaces, channels, playlists, collections, favorites, likes, history, and search results.
• Use right-click table actions to clear completed/failed tasks and open downloaded files/folders.
• Use toolbox functions for audio/video merge, TS segment merge, file merge, and audio extraction.
• If no proxy is configured in the downloader, it follows the system proxy.

Troubleshooting

• No result / low quality: verify membership/login state on Kedou and target platform; for platform-specific list or high-quality content, the user may need a fresh cookie configured locally in the downloader.
• Batch list empty: cookie may be expired, target page may be inaccessible, or the target platform changed its API. Refresh cookie locally and retry.
• Download starts but no file appears: check points/credits, task failure state, duplicate task detection, save path, and disk space.
• Browser download opens playback page: use the player/browser download control.
• Timeout / interrupted parse: wait and retry; some platforms queue or rate-limit parsing.
• Mac install warning: only apply Gatekeeper/quarantine workarounds to a trusted official Kedou installer.

Local Handoff

After Kedou produces files, prefer local CLI tools:

• ffmpeg for media conversion, audio extraction, or merging.
• magick/ImageMagick for thumbnail contact sheets or image post-processing.
• rg, fd, yazi, and fzf for file discovery and review.
• mlx_whisper or whisper-cli when subtitles are absent and transcription is needed.
• Knowledge Hub notes should store durable summaries, file paths, source page families, and processing decisions, not raw credentials or private URLs.

Bilibili subtitle → note (automated recipe)

For a Bilibili video/space URL the repo ships an automated path: scripts/kedou-bili-subs.sh drives this caption page via OpenCLI and prints the downloaded Chinese .srt; pipe it into scripts/forge.sh <srt> --from-text --only both. Full steps + gotchas: references/bilibili-subtitle-to-note.md. Note shape: templates/subtitle-note-template.md (agent fills it; scripts never write the vault).

References

• Read references/kedou-capabilities.md for supported page families, workflows, and OpenCLI-derived source notes.
• Read references/bilibili-subtitle-to-note.md for the validated B站 字幕→笔记 recipe (OpenCLI + Kedou web, filesystem download wait).