The Everything Library — hear-my-book.com

New here? Read GETTING-STARTED.md — the step-by-step publishing guide with the watch-outs. PUBLISHING.md is the pipeline reference, ROADMAP.md the living improvement backlog, CLAUDE.md the working conventions and invariants, and tests/ the suite CI runs on every push.

The library and audio home of the Everything series by Karl Meves — a small static site that showcases every book (read links, EPUB downloads) and serves chapter-by-chapter audio (NotebookLM-generated episodes), plus any slides or infographics that come with them, for:

• Everything That Glows — information is physical
• Everything That Grows — the oldest language
• Everything That Knows — the last question
• Everything That Shows — the weight of looking
• Everything That Goes — book five, in the works (audio previews can be published before the text: upload to media-goes and they appear as extras; when the book is published, fill in its chapters and readUrl in catalog.json)

The player remembers each listener's position per chapter (on-device, localStorage, nothing tracked), shows chapter length and outline, supports play/pause/seek, ±15/30s skips, playback speed, a sleep timer, auto-advance, and drives the lock screen / Bluetooth / headphone controls via the Media Session API. Chapters without audio yet link straight to the book's text, so readers can always choose to read or listen.

How it works

catalog.json      what exists: 4 books × chapters, each with audio/outline/slides
index.html        the player (single file, no dependencies, PWA)
tools/sync.mjs    wires GitHub Release assets into catalog.json

The media lives in GitHub Releases on this repo — one rolling release per book, tagged media-glows, media-grows, media-knows, media-shows. Release assets are free storage, up to 2 GB per file. At deploy time, Netlify's build step (tools/fetch-media.mjs) pulls every file referenced by catalog.json into media/ and serves it same-origin with a real audio MIME type and byte-range support — required because GitHub forces release downloads to application/octet-stream, which iOS Safari refuses to play in an <audio> element. Listening bandwidth therefore flows through Netlify (free tier: 100 GB/month ≈ ~2,600 full 40-minute chapter plays); if traffic ever outgrows that, point the sync's url field at an R2/S3 bucket with free egress — the player only follows the catalog.

Publishing a new episode (the pipeline)

1. Generate the audio in NotebookLM and download the file.
2. Rename it to the convention <book>__<chapter-slug>.mp3, e.g. glows__ch-remembers.mp3. Chapter slugs are the ch-… ids from each book (they're all listed in catalog.json). A slug that isn't a chapter becomes an "extra" — e.g. glows__overview.mp3 for a whole-book Audio Overview. (A single underscore, glows_ch-remembers.mp3, is tolerated for plain audio files — but the double underscore is the canonical form and required for __outline / __slides companions.)
3. Upload it to the book's media release on github.com → Releases → media-<book> → Edit → drag the file into the assets box. (First time: create a release with that tag — media-glows etc.)
4. Run the sync: Actions → Sync catalog from releases → Run workflow (it also runs daily on its own). It reads the file's real duration, updates catalog.json, and commits — Netlify redeploys and the episode is live, linked to its chapter.

Optional companions, same drag-and-drop, matched by filename:

file	shows up as
glows__ch-remembers__outline.txt (or .md)	the chapter's expandable outline
glows__ch-remembers__text.md (or .txt)	a read-along Text tab in the Now Playing view
glows__ch-remembers__slides.pdf	"Slides" button (opens the PDF)
glows__ch-remembers__slide-01.png, -02 …	"Slides" button (inline images)

.m4a, .wav, and friends are accepted too — the sync workflow first normalizes any non-MP3 audio (CI transcodes it to a web-safe 128 kbps MP3 with ffmpeg and uploads it back to the release; the original stays put and the catalog uses the MP3). So upload exactly what NotebookLM gives you. Deleting a chapter's assets from the release and re-running the sync unwires it again — the releases are the source of truth.

Locally, the same sync runs with node tools/sync.mjs (add --dry-run to preview; set GITHUB_TOKEN if you hit API rate limits).

In the car (CarPlay / Android Auto)

Playback uses a real media stream with full Media Session metadata, so when a phone is connected to a car the audio routes as proper media: it ducks for navigation prompts, pauses for calls, shows the book cover, chapter title, and progress on the car's Now Playing screen, and steering-wheel / lock-screen controls work. Start playback on the phone; the car takes it from there.

Web apps cannot appear as apps on the car screen (Apple and Google only allow approved native apps). For car-screen browsing, every book with audio publishes a standard podcast feed (/feed-<book>.xml, generated at deploy by tools/build-feeds.mjs) — add it to Apple Podcasts, Overcast, or any podcast app and the chapters get a native CarPlay / Android Auto interface for free. The feeds can also be submitted to podcast directories whenever that's wanted.

Player URLs

• / — library
• /#glows — one book's chapter list
• /#glows/ch-remembers — deep link that cues a chapter (this is the link format the book sites can use for a "Listen" button per chapter)

Development

No build step. python3 -m http.server in the repo root, open http://localhost:8000. tools/ is stdlib-only (Node 20+): duration.mjs parses MP3/M4A/WAV lengths, make-icons.mjs regenerates the PWA icons. Bump CACHE in sw.js when shipping player changes so installed copies refresh. Audio is intentionally never cached by the service worker — intercepting media range requests breaks seeking.

Publishing

PUBLISHING.md documents the full Errerlabs pipeline — episode loop, adding a new book, EPUB releases, legal defaults, domain setup, and the launch checklist.

Author / publisher

By Karl Meves, published by Errerlabs (errerlabs@gmail.com). Site, books, audio: all rights reserved — see LICENSE and legal.html.