feedbard

feedbard turns the blogs you follow into an audiobook library. It polls a list of RSS/Atom feeds, fetches each new post's full article (Substack API, the feed's own content, or Readability over the page, whichever works), translates it into the language you pick, describes the figures and charts so they are narrated instead of skipped, renders the result to speech, and publishes it as an Audiobookshelf *book* library: one folder per article, per-article cover art, ID3-tagged MP3. There is no WebUI. The container runs the pipeline every CRON_INTERVAL_SECONDS (default 3 hours) and logs what it did. Posts already processed are recorded in a SQLite database under the data path, so a restart never re-narrates the same article. Two paths, kept apart on purpose: /app/data the pipeline's own state - database, rendered episodes, covers, and resumable per-block scratch files. /app/library the