zhf883680/layerwatch repository overview

⁠

⁠LayerWatch

AI-powered 3D print monitoring, failure detection, and timelapse video generation for Bambu Lab printers through Home Assistant.

GitHub⁠ · Documentation⁠ · Releases⁠

⁠Why LayerWatch?

LayerWatch uses AI vision to watch your Bambu Lab print through a Home Assistant camera and detect failures early:

• Spaghetti and failed extrusion
• Clogging or severe material buildup
• Print or support displacement
• Nozzle collisions
• Unknown or uncertain print states

The same captured frames can also be encoded into a timelapse video automatically.

⁠Features

• Works with Bambu Lab printers through Home Assistant
• Uses an existing Home Assistant camera entity
• AI anomaly detection with OpenAI-compatible vision endpoints
• Trigger by layer change or fixed time interval
• Shared capture pipeline for AI analysis and timelapse video
• Automatic MP4 encoding with built-in FFmpeg defaults
• Home Assistant persistent notifications
• Optional Bark push notifications
• Optional Home Assistant-controlled lighting for night captures
• Automatic data retention and cleanup
• Chinese and English web interface
• No RTSP, go2rtc, webhook, or image host configuration required

⁠Supported Architectures

linux/amd64
linux/arm64
linux/arm/v7

⁠Quick Start

No source checkout or compilation is required.

mkdir -p "$HOME/layerwatch/data"

docker run -d \
  --name layerwatch \
  --restart unless-stopped \
  --pull always \
  -p 19091:19091 \
  -v "$HOME/layerwatch/data:/app/data" \
  -e TZ=Asia/Shanghai \
  zhf883680/layerwatch:latest

Open the web interface:

http://<server-ip>:19091

Go to Settings and configure:

• Home Assistant URL and long-lived access token
• Camera, current-layer, and optional print-status entities
• AI endpoint, API key, model, and maximum calls per print
• Layer or interval trigger mode
• Automatic light control with configurable pre-capture delay
• Home Assistant and Bark notifications

Configuration and data are preserved in:

$HOME/layerwatch/data/config.yaml
$HOME/layerwatch/data/monitor.db
$HOME/layerwatch/data/frames/
$HOME/layerwatch/data/videos/
$HOME/layerwatch/data/events/

⁠Screenshots

⁠AI Analysis

⁠Timelapse

⁠Settings

⁠Update

docker pull zhf883680/layerwatch:latest
docker rm -f layerwatch

Run the same docker run command again. The mounted data directory is preserved.

⁠Docker Tags

zhf883680/layerwatch:latest
zhf883680/layerwatch:main
zhf883680/layerwatch:vX.Y.Z
zhf883680/layerwatch:sha-<commit>

⁠AI Cost

Using the example pricing for qwwen-3.7-flash:

• Normal input: CNY 0.2 per million tokens
• Cache-hit input: CNY 0.02 per million tokens
• Output: CNY 0.8 per million tokens

At an estimated 30 AI calls per 3D print, the estimated AI cost is approximately:

CNY 0.0098 per print
≈ 0.98 Chinese cents

See the complete calculation in the GitHub documentation⁠.

⁠Environment Variables

⁠Source Code

GitHub: https://github.com/zhf883680/LayerWatch⁠

If LayerWatch is useful to you, consider giving the repository a star.