Bark!

Listens through your Mac's microphone while you're out, and sends you a Telegram message when your dog barks or whimpers.

Download for macOS

Set up in 10 minutes

You need a Mac and the Telegram app. Nothing else.

Step 1 — Download

Grab the archive above and unzip it (double-click the .zip file if it doesn’t unzip automatically). You’ll get a folder called bark_monitor.

Step 2 — Run the installer

Open the bark_monitor folder and double-click install.command.

A black Terminal window will open and start setting things up. This is normal — just follow what it says on screen:

It may show an Apple popup asking to install “Command Line Tools” — click Install and wait for it to finish, then go back to the Terminal window and press Enter.
It may ask for your Mac password (for Homebrew) — type it and press Enter. (Nothing will show on screen while you type — that’s normal, just type and hit Enter.)
It will download and install Python, TensorFlow, and everything else it needs. This is the slow part — grab a coffee.

Step 3 — Connect Telegram

No need to create a bot or look anything up. Partway through, the installer opens Telegram for you with a chat already started — just tap Start there.

The Terminal window is waiting for that tap and continues on its own once you do — nothing to type or copy.

Step 4 — First launch

The installer finishes by opening Bark!. The first time, macOS will refuse to open it (“unidentified developer”) — this is expected, it’s not from the App Store:

Open Applications in Finder, find Bark!.
Right-click (or Control-click) it → Open → click Open again in the popup.

Bark! will ask for microphone access — click Allow. Without this it can’t hear anything.

Using it

Click Start. Give it 20-30 seconds to load the sound model, then it’ll say “Listening”.
You’ll get a Telegram message when monitoring starts and stops, and whenever it hears barking or whimpering.
Open Logs Folder shows past sessions.

If something’s not working

No sound events ever, even when the dog barks: check System Settings → Privacy & Security → Microphone, and make sure Bark! is switched on there.
App won’t open at all: make sure you right-clicked → Open (see Step 4), not double-clicked.
Installer stops with an error: it’s safe to just double-click install.command again — it picks up where it left off.

Technical details, for the curious

The app is a native macOS SwiftUI client that spawns a Python process as its detection engine: it listens to the microphone via sounddevice and classifies audio in real time with YAMNet (TensorFlow Hub) — recognizing barking, howling, growling, and, most importantly, whimpering (which usually precedes barking in separation anxiety). The Python engine emits structured events to stdout, which the Swift wrapper parses to drive a live “what I’m hearing right now” feed and to control starting/stopping sessions.

Notifications go out over Telegram in two modes. The direct mode talks to your own bot. The shared mode goes through a relay: a separate Docker service on a VPS that’s the only thing holding the real bot token — the client app never sees it. Instead, on first launch it goes through pairing via Telegram: the user taps Start, the relay picks that up via getUpdates, binds the code to their chat_id, and issues a unique notify_token scoped to just that one recipient. A leaked token can’t do anything beyond messaging that same person — unlike the bot token, it doesn’t need to be kept secret.

Distribution is a single install.command that installs Xcode CLT/Homebrew/Python if missing, builds the .venv, compiles the app, and runs the pairing flow — the only thing the user actually does is tap once in Telegram, no terminal, no manual config editing.