Eufy Security Snapshotter with Local Weather Classification

A modular Node.js SDK and utility for capturing snapshots from Eufy Pan/Tilt cameras (including fully custom support for the eufyCam C31 / T817L), rotating the camera to specific preset angles, and running fully offline local AI image inference using ONNX to classify weather conditions.

Key Features

• Object-Oriented SDK Design: Exports a reusable, developer-friendly class that integrates seamlessly into larger applications.
• Persistent Sessions: Automatically saves authentication tokens locally to minimize 2FA prompt frequency.
• Preset Position Control: Moves Pan/Tilt cameras to specific preset indices before taking the capture.
• Efficient Frame Capturing: Establishes local P2P livestreaming, captures a single high-quality frame with FFmpeg, and instantly shuts down the stream. Falling back to Eufy Cloud image URLs if livestreaming is unsupported.
• Local ONNX Weather Inference: Runs a lightweight 5.9MB pre-trained 14-class ResNet50 model offline in under 650ms.
• Re-normalized Prediction Filtering: Restricts predictions to 6 core atmospheric weather classes (cloudy, fogsmog, rain, shine (sunny), rainbow, and snow), re-normalizing their relative confidence to 100% and discarding misleading classifications like lens condensation or dew.

SDK API Reference

Installation

npm install eufy-security-snapshotter

Usage Example

const EufySecuritySnapshotter = require('eufy-security-snapshotter');

async function main() {
    // 1. Instantiate the class with your credentials and config
    const snapshotter = new EufySecuritySnapshotter({
        username: "[email protected]",
        password: "your-password",
        deviceSerial: "T817LT002605061C", // Camera serial number
        presetIndex: 3,                   // Preset to rotate to (optional)
        snapshotsDir: "./custom_snapshots",// Where to save files (optional)
        tokenPath: "./persistent.json"    // Session cache (optional)
    });

    try {
        // 2. Initialize and connect to the Eufy Cloud & Local P2P
        console.log("Connecting...");
        await snapshotter.initialize();

        // 3. Option A: Capture a standard raw image snapshot (returns Buffer)
        const imageBuffer = await snapshotter.takeSnapshot();
        console.log(`Captured image buffer of size: ${imageBuffer.length} bytes`);

        // 4. Option B: Capture snapshot AND run local weather analysis (returns Buffer + Metadata Object)
        const { imageBuffer, metadata } = await snapshotter.takeSnapshotWithWeather();
        console.log(`Captured image buffer of size: ${imageBuffer.length} bytes`);
        console.log(`Detected Weather: ${metadata.weather.label} (${(metadata.weather.confidence * 100).toFixed(2)}%)`);

    } catch (error) {
        console.error("Snapshot failed:", error.message);
    } finally {
        // 5. Cleanly close the session
        await snapshotter.close();
    }
}

main();

Demo Utility

This repository includes a standalone demo utility (demo.js) allowing you to trigger on-demand snapshots and verify your setup.

1. Prerequisites

• Node.js: Latest LTS version.
• FFmpeg: Must be installed on your system path.
  • Amazon Linux / RHEL / Fedora:

    sudo dnf install ffmpeg-free -y

  • Ubuntu / Debian:

    sudo apt update && sudo apt install ffmpeg -y

2. Setup Configuration

1. Clone the repository and install dependencies:

   git clone https://github.com/elijahparker/eufy-security-snapshotter.git
   cd eufy-security-snapshotter
   npm install

2. Copy the environment template:

   cp .env.example .env

3. Open .env and fill in your details:

   EUFY_USERNAME="[email protected]"
   EUFY_PASSWORD="your-eufy-account-password"
   DEVICE_SERIAL="T817LT002605061C"
   PRESET_INDEX=3

3. Run the Demo

To capture a snapshot and run the ONNX weather classification (saves a .jpg and .json in ./snapshots):

node demo.js

To run a raw snapshot only without loading the ONNX model or running classification:

node demo.js --only-snapshot

4. View Saved Weather Logs

A companion .json file is saved with each capture preserving the complete list of normalized class scores:

# Print the results of the newest snapshot weather analysis
cat $(ls -t snapshots/*.json | head -n 1)

Example JSON metadata output:

{
  "timestamp": "2026-06-18T14:01:49.389Z",
  "camera": {
    "name": "Garage",
    "serial": "T817LT002605061C",
    "model": "T817L"
  },
  "weather": {
    "label": "cloudy",
    "confidence": 0.6898,
    "scores": {
      "cloudy": 0.6898,
      "rain": 0.2014,
      "fogsmog": 0.0542,
      "shine": 0.0315,
      "snow": 0.0182,
      "rainbow": 0.0049
    }
  }
}

License

MIT License.