autowhatsapper

A tiny WhatsApp bot that answers one friend's morning greeting for you — with a fresh, AI-written reply every day, so it never reads like a canned message.

When a specific contact sends "Moin", "Moin Moin" or "Guten Morgen", autowhatsapper waits a natural-looking few minutes and replies with a one-off, AI-generated good-morning message. It links to your personal WhatsApp account like WhatsApp Web — no business API, no browser — and is configured entirely through a small built-in web page.

Warning

This uses an unofficial library with your personal account, which is against WhatsApp's Terms of Service. For a low-volume, single-contact personal bot the practical risk is low, but not zero. Don't use it for spam or mass messaging — that's both against the ToS and not what this project is for.

Documentation

• ⚡ 60-second quickstart — clone, install, scan, done
• Configuration — settings, AI vs. static mode, env vars
• Raspberry Pi guide (DEPLOY.md) — run it 24/7 as a systemd service

Features

• Daily-varied replies — each answer is generated via OpenRouter (default model Claude Haiku 4.5), so the wording is different every morning.
• Looks human — replies fire after a random 5–30 minute delay, carry exactly one fitting emoji, and (optionally) only the first greeting per day is answered.
• Strict trigger — only fires when the whole message is a greeting ("Moin" / "Moin Moin" / "Guten Morgen"); "Guten Morgen, wie geht's?" is ignored.
• No file editing — set the API key, target number, model, and persona in a built-in web UI, and scan the WhatsApp linking QR right in your browser.
• Resilient — survives restarts (a pending reply resumes), retries on model errors, and falls back to a second model if the primary one is down.
• Runs 24/7 — ships with a systemd unit for a Raspberry Pi.

How it works

WhatsApp (your phone)
        │  greeting from the target contact
        ▼
   Baileys socket  ──►  greeting matched?  ──►  wait 5–30 min
   (linked device)                                   │
                                                      ▼
                              OpenRouter (Claude Haiku 4.5) generates a reply
                                                      │
                                                      ▼
                                   reply sent back to the contact

Configuration and live status (connection, QR, last reply, pending reply) are served by a minimal vanilla-Node web UI — no framework, no build step.

⚡ 60-second quickstart

git clone https://github.com/klauspet0r/autowhatsapper.git
cd autowhatsapper
npm install
npm start

Then open http://127.0.0.1:8080 and scan the QR with WhatsApp.

Prefer a guided first run? npm run setup walks you through the target number, API key, and persona on the command line — then just npm start and scan the QR.

Quick start (local)

Requires Node.js 18+ (for the built-in fetch).

git clone https://github.com/klauspet0r/autowhatsapper.git
cd autowhatsapper
npm install
npm start

The web UI starts on http://127.0.0.1:8080 (localhost only). Open it, then:

1. Enter your OpenRouter API key and the target number.
2. Save.
3. A WhatsApp QR appears — on your phone go to WhatsApp → Linked devices → Link a device and scan it.
4. Status flips to Verbunden (connected). It now replies on its own.

The session is stored in auth/, so linking is a one-time step. Your settings live in config.json. Both are git-ignored and never leave the machine.

Configuration

All settings are edited in the web UI (no config files to hand-edit):

Setting	Meaning
Target number	Friend's number, country code, no + (e.g. 491701234567)
API key	OpenRouter key from https://openrouter.ai/keys
Model	OpenRouter model slug (default anthropic/claude-haiku-4.5)
Fallback model	Used if the primary model fails after retries
Persona	Free text describing how the bot should sound
Once per day	Reply only to the first greeting each day

The UI's model dropdown lists every OpenRouter model with an estimated per-reply cost, and can filter to free models only.

AI is optional: switch the reply mode to static to answer from your own predefined texts instead — no OpenRouter key needed.

Environment variables

Variable	Default	Purpose
WEB_HOST	127.0.0.1	Bind address for the web UI
WEB_PORT	8080	Port for the web UI

Run on a Raspberry Pi (24/7)

See DEPLOY.md for the full systemd setup: copy the code, install dependencies, register the service, then configure and link via the web UI — reachable either over an SSH tunnel (default, most private) or restricted to your LAN with a firewall rule.

Important

The web UI exposes your API key and the WhatsApp-linking QR, so it must never be reachable from the internet. On a host with a public IP, a firewall rule — not just the bind address — is what keeps it private. DEPLOY.md covers both access modes.

Project layout

File	Role
index.js	Bot core — WhatsApp connection, greeting match, scheduling
web.js	Built-in config web UI (vanilla Node http)
config.js	Loads/saves the web-editable config.json
autowhatsapper.service	systemd unit for the Raspberry Pi
DEPLOY.md	Step-by-step Pi deployment guide

Runtime files — config.json (your key + number), auth/ (the WhatsApp session), and state.json (once-per-day + pending-reply state) — are created at runtime and are git-ignored. Never commit them. config.example.json shows the shape of the generated config.json (placeholders only — the web UI writes the real file for you).

Built with

• Baileys — WhatsApp Web protocol
• OpenRouter — model routing for the reply generation
• qrcode — renders the linking QR

License

MIT © klauspet0r