Tasta mea

IVA DECK

A custom ESP32 macro keyboard with a desktop companion and behavioral recognition

9 mechanical keys · 3 volume controls · LCD + microSD · USB / Bluetooth · Python + PyQt6

Download Windows app · Install V5 · Demo · Get started · Hardware · Protocol · AI experiment · Română

Designed and built by Ioan-Andrei Ivașcu

IVA DECK combines a custom PCB, ESP32 electronics, embedded software, a Windows desktop application, and a behavioral recognition experiment. Nine programmable keys handle shortcuts and media actions, three physical knobs control application volume, and the companion app exposes the display, storage, buzzer, and proximity sensor.

The original academic project was presented in June 2025 at the Faculty of Mathematics and Computer Science, University of Bucharest, as “Tastatură Inteligentă cu Recunoaștere Comportamentală asistată de AI.”

Windows app — V5

Download the Windows installer · Ghid de instalare în română · Source and build instructions

Install IVA-DECK-5.0.0-Setup-x64.exe, pair your deck over Bluetooth or connect USB, then select its port in Conexiune. No separate Python installation is needed. Supports Windows 10 21H2+ / Windows 11 x64.

V5 adds a purple/pink interface, macro labels and assignment dialogs, an editable macro library and three animated potentiometers. The Volum page selects audio targets. Macros, audio targets and the last port are saved per user. AI and SD-image workflows are outside this release. V4 and the historical project documentation remain available.

The V5 application is free software under GPLv3. This first installer is unsigned; see the release notes and SHA-256 checksums.

See it working

Full demonstration with audio — 2 min 8 sec

The recording shows key presses, macro configuration, audio controls, proximity feedback, display interaction, and behavioral analysis. The GIF is a silent excerpt. Open or download the MP4 for the complete recording.

What I built

Desktop companion

Live keys	Application volume

Macros	Display and storage

The application targets Windows, using pycaw and Windows COM APIs for audio-session control. Its interface is in Romanian.

More application screens: proximity and connections

Proximity	Connection settings

Architecture

flowchart LR
  Keys[3x3 matrix] -->|GPIO| ESP[ESP32]
  Knobs[Three knobs] -->|ADC| ESP
  ToF[VL53L0X] -->|I2C| ESP
  ESP -->|SPI| LCD[ST7789]
  ESP <-->|SPI| SD[microSD]
  ESP --> Buzzer[Buzzer]
  ESP <-->|JSON over USB or Bluetooth serial| App[PyQt6 app]
  App --> Macros[OS macros]
  App --> Audio[Windows audio]
  App --> Logs[Local logs]
  Logs --> Features[Feature extraction]
  Features --> RF[Random Forest]

The INMP441 microphone is included in the hardware design. Full host audio streaming and voice control are unfinished, rather than demonstrated features.

From prototype to PCB

PCB layout	Enclosure design

Original hardware iterations are preserved. Start with revision v6 and the hardware guide.

Explore

Project status

The desktop source, AI script, KiCad projects, and peripheral tests are available. A complete historical ESP32 sketch was recovered from the author's development chat and is included with provenance and compatibility notes. It has not been verified as the exact firmware used in the demo; see the firmware status.

The thesis reports approximately 70% behavioral recognition in its experiment. This is a historical reported result, not a newly reproduced benchmark or an authentication guarantee. See the AI guide.

Documentation and media were checked against the supplied materials and repository source. Fresh hardware operation, firmware compilation, and desktop integration have not been revalidated during this documentation update.

Author

Ioan-Andrei Ivașcu · @AndreiasM5

Custom hardware, firmware development, companion software, and behavioral-recognition research for the original academic project.