🗼 WatchTower

The ULTRALATEST Advanced Surveillance Hub & AI Intelligence Console for Frigate NVR.

WatchTower is a comprehensive, real-time surveillance dashboard designed to supercharge your Frigate NVR experience. It combines high-performance live monitoring with intelligent background automation, multi-channel alerts, and Google Gemini AI vision assessment.

(Placeholder for your awesome dashboard screenshot)

🙏 Built On

WatchTower is a console layered on top of other people's excellent work — it doesn't replace or reimplement them:

Frigate NVR by Blake Blackshear — the open-source NVR and object-detection engine that does all the actual camera processing, recording, and detection.
BirdNET-Go by Tomi Hakala — the real-time bioacoustic bird identification engine behind the Bioacoustic Yard Intelligence features.

If you find WatchTower useful, consider starring and supporting those projects directly.

🎯 Finally Fixes Frigate's Parked-Car Problem

Your own car re-triggering as a "new" detection every time light or shadow shifts is one of the most-discussed unsolved annoyances in the Frigate community — zones, masks, object filters inside Frigate itself don't reliably fix it long-term. WatchTower doesn't try to fix it in Frigate. It sidesteps the problem entirely, with two independent mechanisms you can use together — one filters by where a detection is, the other by what Frigate recognizes it as. Both live in the Zones Studio tab.

Exclusion Zones — filter by location

Draw a region directly on a camera's real live frame — around a car in the driveway, a flag, a tree branch, anything that keeps falsely triggering. Any detection centered inside that zone is silently skipped before a notification goes out, independent of whatever Frigate itself concludes about the object's motion. The event still records and still shows up in Review — you just stop hearing about it.

Draw, drag, done: click to place vertices, drag any vertex to reshape, drag inside the shape to move the whole zone — no redrawing from scratch when the car parks a few feet differently.
Per camera, unlimited zones: cover more than one trouble spot per camera, any polygon shape, any number of points.
See it, don't guess: the Snapshot Viewer and 10-Second Playback windows both show your configured zones overlaid on the real footage, so you can check at a glance whether a past event would have been filtered.

The tradeoff to know about: a zone is location-based, not identity-based. If your vehicle needs a large or awkwardly-placed zone to stay covered — say, one that also swallows the entrance to your driveway — an actual unrecognized vehicle parked in that same spot goes just as unnoticed as your own car does. A wide enough zone can trade a false-positive annoyance for a real blind spot. That's exactly what Known Vehicles is for.

Known Vehicles — filter by identity

If you're running Frigate+ with a custom-trained sub-label classifier (Frigate+'s own feature for recognizing a specific vehicle by name, not just "car" generically — see Frigate+'s docs), WatchTower can use that classification directly: list a sub-label name (e.g. Tundra, Rav4) against a camera, and any car detection Frigate identifies as that specific vehicle is skipped — regardless of where in the frame it is.

Because this suppresses by identity rather than location, it doesn't create the blind spot a large exclusion zone can: your own recognized vehicle is silently filtered no matter where it's parked that day, while an unrecognized car — or a stranger's truck — sitting in that exact same spot still alerts normally. No zone to draw, no region of the frame ever goes dark.

Per camera, unlimited vehicles: list every vehicle you have a trained classifier for, per camera.
Exact match required: the name must match your Frigate+ sub-label classifier's name exactly (case-sensitive) — WatchTower doesn't do its own vehicle recognition, it just acts on what Frigate+ already decided.
Complementary, not exclusive: use Known Vehicles for your own recognized vehicles and Exclusion Zones for everything else that isn't identity-based (a flag, foliage, a chronic reflection) — a detection only needs to clear one of them to be skipped.

Find both in the Zones Studio tab (admin-only, since both write to shared notification settings).

✨ Key Features

🚀 Real-Time Tactical Hub

High-Speed Grid: Monitor every camera at once — WebRTC when go2rtc can offer it (real audio+video, no per-camera connection limit), falling back to polled snapshots automatically for anything it can't.
Live Heartbeat: Real-time MQTT connection status with visual "Green Light" confirmation.
Dynamic Telemetry: Live tracking of CPU usage, Coral TPU inference speeds, and system uptime directly from your NVR.
Stability First: De-duplicated event list ensures you see a single, real-time row per detection instead of hundreds of updates.

📧 Intelligent Background Notifications

Persistent Sentinel: Alerts are processed server-side. Receive notifications even when your browser is closed.
Visual Gmail Alerts: Receive high-resolution snapshots of detected objects embedded directly in your emails.
One-Click Action: Every alert (Gmail, Discord, Slack) includes a direct link to "View Event Recording" via your Tailscale or local network.
Multi-Channel Support: Native support for Gmail (SMTP), Discord Webhooks (with image uploads), and Slack Incoming Webhooks.
Per-Camera Control: Mute notifications for individual cameras without affecting recording or the Review feed.

🧠 Advanced AI Filtering & Analysis

Exclusion Zones: Draw a region around a chronic false-trigger spot (a parked car, a flag, a tree branch) — detections centered inside it never reach a notification, regardless of what Frigate itself thinks about the object's motion. See the highlight above.
Known Vehicles: Recognize a specific vehicle by its Frigate+ sub-label and skip it wherever it's parked, without masking off any region of the frame. See the highlight above.
Parked Car Logic: A lighter-weight companion — trusts Frigate's own stationary flag to filter vehicle alerts, no zone drawing or Frigate+ classifier required. Exclusion Zones/Known Vehicles are the more reliable options when that flag gets fooled by changing light/shadow.
Tactical AI Briefs: Integrated with Google Gemini 1.5 Flash to generate human-readable security assessments of events.
Natural Language Search: Find specific events using AI-powered search (e.g., "Show me all the delivery trucks from yesterday morning").

📐 Zones Studio

Exclusion Zones & Known Vehicles: the two parked-car/false-trigger filters described above, in one place — see the highlight near the top of this README for the full explanation of how each works and when to use which.
Admin-only: both write to shared notification settings, so this tab (like the Notifications tab) is hidden entirely for standard user accounts.

🎬 Reliable Clip Playback

Automatic Transcoding: H.265/HEVC event clips (common on newer 4K cameras) are transcoded to browser-compatible H.264 on the fly — Firefox/Chrome can't decode HEVC natively.
Hardware Acceleration: Uses Intel Quick Sync (VAAPI) when available, with automatic fallback to software encoding.
Clip Caching: Transcoded clips are cached, so replaying the same event is instant after the first view.

🐦 Bioacoustic Yard Intelligence

BirdNET-Go Integration: Real-time bird species identification via high-fidelity audio analysis.
Live Audio Sentinel: Listen to your yard in real-time with a built-in frequency spectrogram.
Diversity Report: Automatic population summary of all species visiting your property.
Audio Proof: Play back specific bird song recordings directly from your yard history.
Daily Species Sentinel: Intelligent alerts for the first sighting of each unique species every day, preventing notification fatigue.

🌊 Tidal Intelligence

CHS Predictions: Official Canadian Hydrographic Service tide data (DFO IWLS API) — no API key required.
Multi-Station: Track up to four stations (home, cottage, marina) and switch between them with sub-tabs.
30-Hour Curve: Live SVG tide curve with a "now" marker, interpolated current height, and high/low markers.
Upcoming Tides: High & low table plus a "next high/low in Xh Ym" countdown.
Sun & Moon: Sunrise, sunset, and moon phase computed locally from each station's coordinates.
Tide Alerts: Optional high/low tide notifications a configurable number of minutes ahead, delivered through your existing Gmail / Slack / Discord channels, plus a live in-app banner.

✈️ Live Air Traffic

PiAware / dump1090-fa Integration: Live aircraft positions from your own local ADS-B receiver — no third-party tracking service required.
300nm Range Map: Street, Satellite, and Topographic layers, centered on your home with 50nm range rings and a one-click 2x zoom for the traffic cluster overhead.
Flight Table: Altitude, speed, heading, squawk, and distance from home for every aircraft currently in range.
Aircraft Detail: Click any plane for its photo, type, registration, and origin/destination — sourced from free community APIs (adsbdb.com, planespotters.net), with optional departure/arrival times via a free registered OpenSky Network API client.

🌤 Weather

Current Conditions: Temperature, feels-like, humidity, wind, pressure, and precipitation for your home location.
7-Day Forecast: Daily highs/lows, precipitation chance, and sunrise/sunset — no API key required (Open-Meteo).

🛠 Installation & Setup

Prerequisites

A running instance of Frigate NVR.
Node.js v22+ installed on your server.
ffmpeg (and ffprobe) installed and on PATH — required for clip transcoding and BirdNET-Go live audio.

1. Clone & Install

git clone https://github.com/jchisholm59/WatchTower.git
cd WatchTower
npm install

2. Configure Secrets

Create your environment files in the root directory (use .env.example as a template).

Important

To prevent AI development environments from overwriting your custom settings, this project uses two environment files:

.env: Managed automatically by the IDE (stores your GEMINI_API_KEY).
guardian.env: Created by you for all other secrets (Gmail, Slack, Discord, etc.).

Example guardian.env:

GMAIL_USER=your-email@gmail.com
GMAIL_PASSWORD=your-google-app-password
GMAIL_RECIPIENT=your-alerts-recipient@gmail.com
# GEMINI_API_KEY is handled in .env

3. Run for Production (Recommended)

We recommend using PM2 to keep the sentinel running 24/7 in the background:

npm run build
pm2 start dist/server.cjs --name watchtower

4. Running with Docker (Recommended)

You can also run WatchTower using Docker, which is the recommended way for production deployment. It also supports Intel Quick Sync (VAAPI) hardware-accelerated clip transcoding when /dev/dri is available.

Quick Start:

Configure environment:
- The IDE will manage your GEMINI_API_KEY automatically in the .env file.
- Create a file named guardian.env for your custom secrets (use .env.example as a template).
Start the container:
```
docker compose up -d --build
```

The application will be available at http://localhost:8100 (configurable via the PORT environment variable in docker-compose.yml).

Persistent Data: Docker will automatically create a volume to persist your settings:

guardian_data: Persists notification_settings.json, mqtt_config.json, users.json (login accounts), and the transcoded clip cache in the /app/data directory inside the container.

Troubleshooting & Maintenance:

View logs: docker compose logs -f
Restart: docker compose restart
Full Reset (Wipes all settings, credentials, and login accounts): If you want to perform a truly clean install and wipe all persisted settings from the Docker volume — this also deletes every account, so the next start re-seeds the default admin/watchtower login:
```
docker compose down -v
docker compose up -d --build
```

5. Enable Live Audio in the Camera Detail View (WebRTC)

Opening a camera's expanded "Inspect & Detail" view streams over WebRTC via Frigate's embedded go2rtc — the grid also uses WebRTC when a camera supports it, but every tile there is muted by default (nobody wants 7 cameras talking over each other at once). The detail view is where you actually get sound, with a mute/unmute toggle. To get video and audio there, one small addition to Frigate's own config.yml is usually required:

go2rtc:
  streams:
    ...(your existing camera entries)...
  webrtc:
    candidates:
      - YOUR_FRIGATE_HOST_IP:8555

Add webrtc: as a sibling of your existing go2rtc.streams key — not a second top-level go2rtc: block elsewhere in the file. YAML doesn't merge duplicate top-level keys; the last one silently wins, which would wipe out every camera restream you've already defined under streams:. Restart Frigate after saving.

Why this is needed: without it, go2rtc has to guess its own reachable address for WebRTC's ICE negotiation, which frequently fails on a Docker host with more than one network interface — the connection succeeds, but no media ever actually reaches the browser. This is a Frigate/go2rtc configuration detail, not something WatchTower can work around in code.

If you get audio but no video: that's a separate, unrelated issue — most browsers' WebRTC stack can negotiate H.265 but never actually decode it, so a camera whose stream happens to be H.265 will play audio while video silently never renders. WatchTower already works around this automatically by preferring an H.264 stream when one of the camera's configured ffmpeg inputs offers it — if a camera only has an H.265 source available at all, there's currently no fix short of changing that camera's stream encoding.

🔐 Login & Accounts

WatchTower requires signing in — there's no way to reach the dashboard, live feeds, or any API route without an account.

First run: if no accounts exist yet, WatchTower seeds a default one automatically:

Username: admin
Password: watchtower

Log in with that immediately and change the password (top-right Account menu → Change My Password) — don't leave it at the default, especially if this server is reachable from outside your LAN.

Two roles:

Admin — full access, including adding/removing accounts and changing shared system configuration: Frigate servers, MQTT, and the Notifications settings screen (which covers Gmail/Slack/Discord/Filters, BirdNET-Go, Tides, Flights/PiAware, and Weather).
Standard — identical use of the app otherwise (live feeds, event review, Birds/Tides/Flights/Weather tabs, AI search, etc.), but can't touch any of the configuration above — it's hidden from the UI entirely and rejected server-side if requested directly. Good for other household members who should be able to use WatchTower without being able to break it.

Managing accounts: any admin can add more accounts (admin or standard) from the Account menu, and reset another user's password if they forget it — there's no email-based recovery. Every logged-in user can change their own password from the same menu.

Accounts are stored in <data dir>/users.json (~/.frigate-guardian/users.json, or /app/data/users.json under Docker) — bcrypt-hashed passwords, never committed to git, never leave your server.

🐦 BirdNET-Go Setup

To enable bird song identification, go to Notifications -> BirdNET-Go in the dashboard:

Enable Integration: Toggle the switch to ON.
MQTT Broker: Provide the IP of the broker BirdNET-Go is publishing to.
MQTT Topic: Set to your sightings topic (default: birdnet-sightings).
Web URL: Enter your BirdNET-Go web interface address (e.g., http://192.168.2.210:8080) to enable audio clip playback.
Live Audio: Enter the RTSP URL of your yard microphone to enable the live spectrogram.
Daily Alerts: Toggle "Daily Species Sentinel" to receive notifications for the first detection of each species every day.

🌊 Tides Setup

Tide predictions come from the public DFO / Canadian Hydrographic Service IWLS API — no key or account needed. In the dashboard, go to Notifications -> Tides:

Enable Integration: Toggle the switch to ON.
Add Stations: Search by name or 5-digit CHS code (e.g. Halifax, Digby, 00490) and add up to four.
Units: Choose metres or feet.
Tide Alerts (optional): Toggle on, set how many minutes ahead to notify, pick high and/or low tide, and choose which channels carry them. Alerts reuse the webhook URLs / SMTP credentials from the Gmail, Slack, and Discord tabs.

Predictions are cached server-side for 10 minutes. The Tides tab shows the curve, upcoming tides, and sun/moon panel for each configured station. Requires outbound HTTPS to api-iwls.dfo-mpo.gc.ca.

✈️ Flights Setup

Live air traffic requires a local PiAware or dump1090-fa ADS-B receiver on your network (a Raspberry Pi + RTL-SDR dongle is the common setup). In the dashboard, go to Notifications -> Flights:

Enable Integration: Toggle the switch to ON.
PiAware Receiver Address: Just the IP (e.g. 192.168.1.x) — WatchTower fills in the standard /skyaware/data/aircraft.json path. A full URL also works for non-default installs. Use Test Connection to confirm it's reachable.
Home Coordinates: Enter your latitude/longitude to center the map and range rings.
OpenSky Network (optional): For departure/arrival times, register a free API client at opensky-network.org and paste the Client ID/Secret in. Without it, you still get aircraft photos and routes for scheduled flights, just no timing data.

The Flights tab polls the receiver every few seconds — no data leaves your network except the on-demand aircraft photo/route lookups (adsbdb.com, planespotters.net, optionally OpenSky) when you click a specific plane.

🌤 Weather Setup

Uses Open-Meteo — free, no API key or account needed. In the dashboard, go to Notifications -> Weather:

Enable Integration: Toggle the switch to ON.
Home Coordinates: Enter your latitude/longitude (the same values as Tides/Flights, if configured).

The Weather tab refreshes every 10 minutes, matching the server-side cache.

📥 Sorting Alerts in Gmail (Optional)

Every alert email WatchTower sends uses a predictable subject line: 🚨 [Frigate Alert] <OBJECT> detected on <camera> (<LEVEL> Threat). That's enough structure to auto-sort alerts out of your inbox into labels by detected object type, without touching the app itself — all Gmail-side.

Label structure: a parent Frigate label (catches every alert, archived out of the inbox) plus four sub-labels: Frigate/Car, Frigate/Person, Frigate/Bird, and Frigate/Wildlife for everything else — non-bird BirdNET-Go sounds and any non-car/non-person object Frigate's own video detection reports (dog, cat, moose, raccoon, bicycle, whatever the model supports).

How Bird matching works — anchored on the camera name, not the species name. Car and Person alerts (from Frigate's own object detection) literally contain the words "CAR"/"PERSON" in the subject, so those are easy exact matches. Bird alerts, via the BirdNET-Go integration above, instead carry a specific species name (e.g. "BARRED OWL", "BLACK-CAPPED CHICKADEE") rather than the word "bird" — so matching by species name would mean an ever-growing keyword list. Instead, the Bird rule matches on ESP32-C in the subject's camera slot: every BirdNET-Go alert comes from that yard microphone's camera entry (ESP32-C6 on this setup — ESP32-C also covers other C-series boards like C3), and Car/Person alerts never do (they come from the porch/driveway/yard video cameras), so this is a reliable positive match rather than a guess.

That same yard microphone occasionally picks up other vocalizing wildlife too — real detections so far have included coyote, frog (green frog, wood frog), squirrel, and chipmunk chatter, and bat is plausible as well since BirdNET-Go's classifier isn't birds-only. Those get excluded from Frigate/Bird by name (filter 4's exclusion list).

Wildlife is the true catch-all, not just a positive species list. Gmail filters aren't evaluated top-down or exclusively — every filter that matches a message fires, and there's no "stop processing" action, so a naive "Wildlife = contains COYOTE/FROG/etc." filter would double-label those messages as both Bird and Wildlife (they also contain ESP32-C). Filter 5 instead matches the logical complement of Car/Person/Bird directly: not Car, not Person, and (not from ESP32-C at all, or one of the known non-bird ESP32-C sounds). The "not from ESP32-C at all" branch is what catches a Frigate video detection Filter 2/3 didn't recognize (a DOG or CAT alert from the porch/driveway/yard cameras, say) — those aren't birds and aren't Car/Person, so they land in Wildlife rather than sitting unsorted under just the parent Frigate label. If a new non-bird ESP32-C sound starts showing up in Frigate/Bird instead of Frigate/Wildlife, add it to both filter 4's exclusion list and filter 5's OR-group below, then re-import.

Setup — Gmail supports bulk filter import, which is much less painful than the "Create filter" dialog for queries this long:

Save the XML below as a file (e.g. frigate-gmail-filters.xml).
In Gmail: gear icon → See all settings → Filters and Blocked Addresses tab → scroll down → Import filters.
Choose the file, review the preview Gmail shows, then Create filters.

<?xml version='1.0' encoding='UTF-8'?>
<feed xmlns='http://www.w3.org/2005/Atom' xmlns:apps='http://schemas.google.com/apps/2006'>
  <title>Mail Filters</title>

  <!-- 1. Catch-all: every Frigate alert gets the Frigate label and is archived out of the inbox -->
  <entry>
    <category term='filter'></category>
    <title>Mail Filter</title>
    <content></content>
    <apps:property name='hasTheWord' value='subject:("Frigate Alert")'/>
    <apps:property name='label' value='Frigate'/>
    <apps:property name='shouldArchive' value='true'/>
  </entry>

  <!-- 2. Car alerts -->
  <entry>
    <category term='filter'></category>
    <title>Mail Filter</title>
    <content></content>
    <apps:property name='hasTheWord' value='subject:("Frigate Alert") subject:(CAR)'/>
    <apps:property name='label' value='Frigate/Car'/>
  </entry>

  <!-- 3. Person alerts -->
  <entry>
    <category term='filter'></category>
    <title>Mail Filter</title>
    <content></content>
    <apps:property name='hasTheWord' value='subject:("Frigate Alert") subject:(PERSON)'/>
    <apps:property name='label' value='Frigate/Person'/>
  </entry>

  <!-- 4. Bird alerts: anchored on the ESP32-C camera name (BirdNET-Go's yard mic),
       minus the non-bird sounds BirdNET-Go's audio classifier realistically
       vocalize-IDs (coyote, frog, bat, squirrel, chipmunk chatter), not a
       generic wildlife-camera list, since this is audio classification, not
       a trail cam. Add more -subject:(...) exclusions if a new non-bird
       species type shows up, then re-import. -->
  <entry>
    <category term='filter'></category>
    <title>Mail Filter</title>
    <content></content>
    <apps:property name='hasTheWord' value='subject:("Frigate Alert") subject:(ESP32-C) -subject:(COYOTE) -subject:(FROG) -subject:(BAT) -subject:(SQUIRREL) -subject:(CHIPMUNK)'/>
    <apps:property name='label' value='Frigate/Bird'/>
  </entry>

  <!-- 5. Wildlife: the true catch-all, not Car, not Person, and (not from
       the ESP32-C mic at all, OR one of the known non-bird ESP32-C sounds).
       That first branch is what catches a Frigate *video* detection that
       isn't Car/Person (dog, cat, moose, raccoon, whatever the object model
       supports), without it, those would sit unsorted under the parent
       Frigate label. Keep the OR-group in sync with filter 4's exclusion
       list if either one changes. -->
  <entry>
    <category term='filter'></category>
    <title>Mail Filter</title>
    <content></content>
    <apps:property name='hasTheWord' value='subject:("Frigate Alert") -subject:(CAR) -subject:(PERSON) {-subject:(ESP32-C) subject:(COYOTE) subject:(FROG) subject:(BAT) subject:(SQUIRREL) subject:(CHIPMUNK)}'/>
    <apps:property name='label' value='Frigate/Wildlife'/>
  </entry>

</feed>

Gmail matches/creates labels by name on import, so re-importing after editing the exclusion list won't create duplicate labels — it just updates the filter's match criteria.

🔒 Security & Privacy

Login Required: See Login & Accounts above — every route requires a signed-in session, with admin/standard roles separating "use the app" from "reconfigure the app."
Local First: Your passwords, integration credentials, and account list are stored locally in ~/.frigate-guardian (or /app/data under Docker) and never uploaded to the cloud.
Encrypted Streams: Supports HTTPS and secure WebSocket (WSS) for camera intercepts.

🏗 Built With

Frontend: React, Tailwind CSS, Lucide Icons, Framer Motion, Leaflet.
Backend: Node.js, Express, MQTT.js, Nodemailer, FFmpeg.
AI: Google Gemini 1.5 Flash.
Data: Frigate NVR by Blake Blackshear, BirdNET-Go (MQTT) by Tomi Hakala, DFO / Canadian Hydrographic Service IWLS (tides), OpenStreetMap / Esri / OpenTopoMap (map tiles), adsbdb.com, planespotters.net and OpenSky Network (flight & aircraft lookups), Open-Meteo (weather).

Created with ❤️ for the Frigate NVR Community.

Name		Name	Last commit message	Last commit date
.artifacts/fcf2a4ca-4a83-4d0b-a265-0b51c0034c84		.artifacts/fcf2a4ca-4a83-4d0b-a265-0b51c0034c84
.claude		.claude
public		public
scripts		scripts
src		src
.dockerignore		.dockerignore
.env.example		.env.example
.gitattributes		.gitattributes
.gitignore		.gitignore
Dockerfile		Dockerfile
GITHUB_INSTRUCTIONS.md		GITHUB_INSTRUCTIONS.md
README.md		README.md
ROLLBACK_NOTES.md		ROLLBACK_NOTES.md
SERVER_SETUP.md		SERVER_SETUP.md
bun.lock		bun.lock
docker-compose.yml		docker-compose.yml
flights.ts		flights.ts
index.html		index.html
metadata.json		metadata.json
package-lock.json		package-lock.json
package.json		package.json
server.ts		server.ts
tides.ts		tides.ts
tsconfig.json		tsconfig.json
vite.config.ts		vite.config.ts
weather.ts		weather.ts

Navigation Menu

Repository files navigation

🗼 WatchTower

🙏 Built On

🎯 Finally Fixes Frigate's Parked-Car Problem

Exclusion Zones — filter by location

Known Vehicles — filter by identity

✨ Key Features

🚀 Real-Time Tactical Hub

📧 Intelligent Background Notifications

🧠 Advanced AI Filtering & Analysis

📐 Zones Studio

🎬 Reliable Clip Playback

🐦 Bioacoustic Yard Intelligence

🌊 Tidal Intelligence

✈️ Live Air Traffic

🌤 Weather

🛠 Installation & Setup

Prerequisites

1. Clone & Install

2. Configure Secrets

3. Run for Production (Recommended)

4. Running with Docker (Recommended)

5. Enable Live Audio in the Camera Detail View (WebRTC)

🔐 Login & Accounts

🐦 BirdNET-Go Setup

🌊 Tides Setup

✈️ Flights Setup

🌤 Weather Setup

📥 Sorting Alerts in Gmail (Optional)

🔒 Security & Privacy

🏗 Built With

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

🗼 WatchTower

🙏 Built On

🎯 Finally Fixes Frigate's Parked-Car Problem

Exclusion Zones — filter by location

Known Vehicles — filter by identity

✨ Key Features

🚀 Real-Time Tactical Hub

📧 Intelligent Background Notifications

🧠 Advanced AI Filtering & Analysis

📐 Zones Studio

🎬 Reliable Clip Playback

🐦 Bioacoustic Yard Intelligence

🌊 Tidal Intelligence

✈️ Live Air Traffic

🌤 Weather

🛠 Installation & Setup

Prerequisites

1. Clone & Install

2. Configure Secrets

3. Run for Production (Recommended)

4. Running with Docker (Recommended)

5. Enable Live Audio in the Camera Detail View (WebRTC)

🔐 Login & Accounts

🐦 BirdNET-Go Setup

🌊 Tides Setup

✈️ Flights Setup

🌤 Weather Setup

📥 Sorting Alerts in Gmail (Optional)

🔒 Security & Privacy

🏗 Built With

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages