A real-time chat that signs on like a 1998 dial-up service — full modem handshake, "Welcome!" / "You've Got Mail!"-style voices, buddy-list chimes, yellow IM pop-ups, and chat rooms ranging from the classic (Lobby, Music, Gaming, Coding) to the modern wing (Memes, AI, Crypto). Branded as thebardchat, the main chat for the bardchat ecosystem.
Built on Cloudflare Workers + Durable Objects so any two browsers anywhere on the planet land in the same room with sub-second sync.
┌──────────────────────────────────────────────────────────────┐
│ Browser │
│ index.html → signon layer ▶ boot layer ▶ chat layer │
│ │ │
└──────────────────────────────────────────────────┼───────────┘
▼ WebSocket
/ws/:room ┌──────────────────┐
/dm/:a/:b │ Cloudflare Worker│
│ ↓ idFromName │
│ Room (DO) │
│ • hibernatable WS│
│ • member list │
│ • last-50 history│
└──────────────────┘
npm install
npm run fetch-sounds # best-effort; see public/sounds/README.md if any miss
npm run dev # http://localhost:8787wrangler dev serves the static assets from public/ and runs the Worker + Durable Object locally. Open the URL in two browsers and sign on with different screen names to verify multi-user chat.
Durable Objects require the Workers Paid plan ($5/mo).
npx wrangler login
npx wrangler deployThe first deploy runs the v1 migration that creates the Room SQLite DO class. After that, wrangler deploy is idempotent.
| Path | Purpose |
|---|---|
worker/src/index.ts |
Router: WS upgrade for /ws/:room and /dm/:a/:b; everything else → static assets |
worker/src/room.ts |
The Room Durable Object — hibernatable WebSockets, member list, last-50 history |
public/index.html |
Single-page app: sign-on → dial-up → chat |
public/styles/aol.css |
Win95 chrome (titlebars, bevels, buttons, running-man) |
public/styles/chat.css |
Chat-room layout, buddy list, DM windows, tabs |
public/scripts/sounds.js |
Audio preloader + autoplay-unlock helper |
public/scripts/signon.js |
Sign-on form, saved names |
public/scripts/boot.js |
Dial-up sequence + welcome/mail audio |
public/scripts/chat.js |
Room state, WS lifecycle, slash commands |
public/scripts/dm.js |
IM pop-up windows (draggable, multi-instance) |
scripts/fetch-sounds.* |
Best-effort downloader for the four audio clips |
Rooms are keyed by name — the Worker resolves /ws/lobby to env.ROOM.idFromName("room:lobby") so every visitor lands in the same Durable Object. The valid room set is fixed: lobby, music, gaming, coding, memes, ai, crypto.
DM windows route to env.ROOM.idFromName("dm:<a>:<b>") where the names are lowercased and sorted, so both sides land on the same DO regardless of who opened the IM first.
In the chat input:
/me <action>— emote (* CoolDude1998 waves)/join <room>— join one of the seven rooms/leave(or/part) — leave the current room/rooms— open the chat-room finder/reconnect— close and reopen the WS for the current room/help— list commands
Double-click any name in the buddy list or the "People Here" panel to open a DM window.
See public/sounds/README.md. Place four MP3s in public/sounds/ (modem.mp3, welcome.mp3, youve-got-mail.mp3, buddy-in.mp3) or run npm run fetch-sounds and let it try Internet Archive. Missing files are treated as silence so the boot sequence always completes.
- Single Worker + DOs keeps cost flat. With hibernatable WebSockets you only pay duration when a room is actively passing messages.
- No build step for the frontend — vanilla HTML/CSS/JS hits the browser directly. Easier to tweak in DevTools and faster to iterate.
- Single-page app (not two HTML files) so the SIGN ON click carries autoplay credit into the modem audio. Mobile Safari and iOS Chrome are the strict ones; this works around them.
MIT for the code in this repo. The look and sounds are a tribute to the 1998 dial-up era; any third-party audio used during local dev belongs to its respective owner and is not redistributed by this repo.