Release v3.12.0 Pro • 09.10.2026

Autonomous Energy Telemetry & Predictive No-Frost Diagnostics

A complete self-hosted IoT hardware-software complex running on a $15 Android TV Box (Termux/Linux) and Tuya smart plug for real-time power monitoring, chamber temperature modeling without drilling holes, blackout compressor protection, hourly hardware calibration, and smart Telegram alerts with Dual-Tier AI.

Appliance Telemetry (Nominal Cooling Duty Cycle)
Samsung CoolTech Dynamic No-Frost • Host: 192.168.0.103:8088
100% LAN SOCKET (3s)
DUAL AI ACTIVE
Active Power
124.6 W
Cooling (compressor active)
Mains Voltage
218.4 V
Safe nominal range
Duty Cycle (КРВ)
0.38 DC
Energy-efficient ratio
Freezer Chamber
-18.9 °C
Thermodynamic model
Today's Energy: 0.984 kWh ($0.03)
Next Defrost Cycle: ~3h 20m
Energy Calibration (XX:05): 99.9% match
Socket Anti-Stall: Active Guard
100% LAN
Direct smart plug polling without Tuya Cloud quota
±0.5 °C
Thermodynamic model accuracy without chamber probes
5 min
Hardware compressor safety delay against power flickers
XX:05
Hourly hardware calibration with smart plug metering chip
3 Watts
Ultra-low server consumption on Android TV Box (24/7)

Engineering Solutions for Total Appliance Control

The ecosystem continuously tracks electrical parameters, models thermodynamics, and protects the compressor during power interruptions.

Dual-Tier AI Architecture

Decoupling fast deterministic intent routing from deep generative reasoning for maximum autonomy and reliability.

SYSTEM 1 • DETERMINISTIC NLP

TypeSafe AI Jev

Sub-second (~80ms) local intent routing, duty cycle scoring, and mechanical wear evaluation with zero external library overhead.

  • ✓ Deterministic classification ("how's the fridge?", "how much spent today?")
  • ✓ Compressor wear assessment and duty cycle balance calculations
  • ✓ Zero CPU overhead on the 3-Watt Android TV Box server
  • ✓ Complete offline operation without requiring external LLM APIs
Latency: ~80 ms • 100% Deterministic Reliability
SYSTEM 2 • REFRIGERATION EXPERT

Google Gemini Flash

Conversational refrigeration engineer grounded in Samsung RT34MB factory specs and live telemetry with 5-model cascading fallback.

  • ✓ Answers engineering questions ("why are side walls warm?", "safe loading limits?")
  • ✓ Food storage optimization and temperature zone recommendations
  • ✓ Resilient 5-tier model fallback (from flash-lite down to 3.6-flash)
  • ✓ Detects thermal anomalies in R-134a heat dissipation curves
High Availability: 5 Models • Cascading Fallback

Refrigerator Door Kiosk for $0

A retired 2012 iPad transformed into a smart kitchen telemetry panel powered by an ultra-low power Android TV Box.

Vintage iPad 3 Retina running Kiosk UI on Samsung RT34MB refrigerator door
Real-World Home Installation

Apple iPad 3 Retina + Samsung RT34MB

A retired iPad 3 (Retina Display, iOS 9.3.6, Model MD369KS/A), unable to run modern bloated web applications, finds a permanent second life with our pure ES5 table-based interface running 24/7 with zero lag and instant wake.

  • Kiosk Tablet: Apple iPad 3 Retina (iOS 9.3.6)
  • Mounting: Magnetic adhesive on freezer door
  • Frontend Stack: Pure ES5 + CSS3 (<50ms render)
  • 24/7 Microserver: H96 Max RK3318 TV Box (3W)
  • Sensor Polling: Tuya Wi-Fi 16A (LAN Direct 3s)
Open iPad Kiosk Open Desktop Dashboard

Hardware Stack & Data Flow Topology

Robust local-first infrastructure operating autonomously without mandatory cloud dependencies.

STEP 1
Samsung Refrigerator
RT34MBTS • SK170K-T1U R-134a • Power Draw 120–160 W
STEP 2
Tuya Smart Plug
16A Smart Plug • TinyTuya v3.5 • Local Wi-Fi Socket 6668
STEP 3
H96 TV Box (Server)
Rockchip RK3318 Cortex-A53 • Termux + Python 3.12 + SQLite
STEP 4
Clients & Telegram
Door-Mounted iPad • Desktop Web UI • @my_fridge_rt34_bot
Network Policy Routing: The Android TV Box Linux kernel enforces static interface routing for the local router subnet (192.168.0.0/24) via eth0. This guarantees jitter-free smart plug telemetry even if system-wide VPN tunnels or proxies are active on the media box.
Telegram Bot Avatar
Samsung Fridge Monitor
@my_fridge_rt34_bot
Open in Telegram
BLACKOUT RESTORATION ALERT EXAMPLE
220V MAINS POWER RESTORED!

• Outage Start: 17:12
• Outage End: 22:49
• Outage Duration: 5h 38m (20,267 sec)
• Estimated Temp: Freezer reached ~-10°C

Compressor safety delay observed. Standard monitoring resumed.
Interactive Quick Commands:
Status
Weekly Report
Today
Reconcile
Audit
Help

Technical Hardware Specifications

Parameter Specification
Appliance Model Samsung RT34MBTS (CoolTech Dynamic No-Frost, 340 L)
Compressor Model Samsung SK170K-T1U (hermetic reciprocating, PTC starter)
Refrigerant & Charge R-134a, precise factory spec 160 g
Compressor Power 115–135 W (steady-state boiling nominal)
Defrost Heater 155–165 W (active melting 20–28 min every 9–14 h)
Server Host H96 Max TV Box (Rockchip RK3318 Cortex-A53 64-bit, 2GB RAM, 3W)
Runtime Environment Linux (Termux Environment) + Python 3.12 (Flask + TinyTuya LAN)
Database Engine SQLite3 WAL with timestamp indexes (>140,000 readings)
AI Co-Pilots TypeSafe AI Jev (~80ms) + Google Gemini Flash (5-tier fallback cascade)
Kiosk Tablet Apple iPad 2 / iPad 3 / iPad 4 (legacy iOS 8.4 – 12 pure ES5 support)

Deployment on Android / Linux

Deploy the complete server in minutes inside Termux on an Android TV Box, Raspberry Pi, or any Linux/Windows PC.

1. Clone Repository & Install Dependencies:

git clone https://github.com/ALEVOLDON/SmartFridge-Telemetry-Kiosk.git
cd SmartFridge-Telemetry-Kiosk
pip install -r requirements.txt
cp config.example.json config.json

2. Launch Telemetry Server:

# Standard foreground server launch:
python auto_monitor.py

# Or 24/7 background daemon in Termux:
bash scripts/h96_start_fridge.sh

The web dashboard will automatically be available at port 8088: http://<SERVER_IP>:8088/ (for vintage iPad tablet: /ipad).