Skip to content

jev-windows-voice

Electron app for controlling Windows 10/11 by Turkish or English speech, using OpenAI Realtime, Jev decisions, UI Automation and Playwright.

Projectprototype

Overview

Jev Windows Voice is an Electron application that controls a Windows 10 or 11 computer through natural Turkish or English speech. Full-duplex conversation uses OpenAI Realtime models, with local Whisper as an option; typed decisions come from typesafe/jev-1.13 via OpenRouter's Decisions API. Microsoft UI Automation through pywinauto is preferred for acting on applications, with an optional screenshot and vision fallback for interfaces UIA cannot reach, and Jev plus Playwright handle browser automation.

Capabilities include app and window management, UI element selection, typing, keyboard, mouse, scrolling and Windows search. Safety measures listed in the README: no arbitrary shell or terminal tool, desktop tools only work after the user enables computer control, every UI change invalidates the previous snapshot, coordinate clicks are avoided when UIA is available, and API keys and audio are never logged. Keys are encrypted with Electron safeStorage backed by Windows Credential Manager. Global shortcuts toggle listening and trigger an emergency stop. It builds NSIS installers and a portable exe and derives from the open jev-mac-voice and jev-voice-browser projects. The README is in Turkish.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich UI element or action fulfils the spoken command?(inferred)—

How it works

  1. Speech captured via OpenAI Realtime or local Whisper
  2. Jev selects action and target from UIA snapshot or browser page
  3. Execute via pywinauto or Playwright; vision fallback if needed
  4. Snapshot invalidated after every UI change; emergency stop available

Strengths

  • No arbitrary shell execution
  • Encrypted key storage
  • Emergency stop shortcut
  • Installer and portable builds

Caveats

  • Depends on OpenAI Realtime for conversation
  • Windows only; README in Turkish

Getting started

npm.cmd ci, create a Python 3.12 venv, pip install -r native\requirements-windows.txt, then npm.cmd run desktop.

At a glance

Type
Project
Status
prototype
Language
TypeScript
Runtime
Electron + Python bridge
Has code
yes

Use cases and integrations

How curated lists describe it

  • Türkçe ve İngilizce doğal konuşmayla Windows 10/11 bilgisayar kontrolü: OpenAI Realtime, local Whisper, Jev, UI Automation ve Playwright.

    hellogumbo/awesome-jev

Frequently asked questions

Which languages can I speak?

Turkish or English.

Can it run shell commands?

No, there is no arbitrary shell or terminal tool.

How is Jev accessed?

As typesafe/jev-1.13 through OpenRouter's Decisions API.

Related projects

  • RAM-efficient Rust coding-agent harness, listed as using Jev decisions for memory pruning, browser navigation and voice routing.

    ServiceproductionRust
  • macOS menu-bar Swift app that turns on-device transcribed speech into Jev typed decisions and macOS actions, confirming hard-to-undo steps.

    AppbetaSwift
  • Lets coding agents run whole goals in your signed-in Chrome on macOS in background tabs, asking Jev or Claude Haiku only for real choices.

    Projectbeta
  • Computer-use platform (Cua Spaces, Driver, Lume) with CUA-S1 small decision models and a preview jev-use example for bounded actions.

    ProjectbetaPython
  • Local stdio MCP server that executes whole browser and macOS goals with a Jev-driven observe-decide-act loop.

    MCP serverbetaTypeScript
  • Browser Use's browser agent where Jev picks an operation and page element in one request and a small LLM only writes text to type.

    LibrarybetaPython