Building Vision Assistant: A Custom Desktop AI Tool Built with Electron

Chhavi Gupta — Sun, 27 Sep 2026 10:35:25 +0000

This is a submission for the MLH x DEV Writing Challenge

What I Built

I built Vision Assistant v1.0.0, a custom desktop AI vision tool created from scratch using Electron.

What it does: It brings instant, system-wide AI assistance directly to your desktop environment without the friction of constant context-switching. It features global shortcut activation (Ctrl + Shift + Space), multi-screen capture capabilities, real-time chat & vision analysis, and memory persistence across sessions.
The Inspiration: I wanted to eliminate the workflow friction of switching between browser tabs and separate tools when analyzing multiple screen areas or seeking quick AI assistance during development.
What I Learned: Building a desktop app from scratch using Electron taught me a lot about handling system-level global shortcuts, managing multi-screen capture streams, and packaging a clean Windows executable using electron-builder.

Demo

You can check out the setup guide video and walkthrough here:
Vision Assistant Setup Guide Video

You can also download the latest Windows standalone release package directly from the GitHub repository:
GitHub Repository & Releases
release link for windows user

Partner Technologies

For this project, I focused heavily on core web and desktop ecosystems using JavaScript (69%) and HTML (31%), powered by Electron to bundle a seamless standalone Windows application with a dark/purple customized UI theme optimized for desktop environments.

Hackathon Experience

Building Vision Assistant for this challenge was an incredible journey of pushing desktop development boundaries. Bringing an idea from local scripts to a fully packaged, global shortcut-enabled desktop application made it a deeply rewarding learning experience!

Say Hello to Vision Assistant: Your Instant, Shortcut-Triggered Desktop AI Tool

Chhavi Gupta — Tue, 22 Sep 2026 18:28:49 +0000

``This is a submission for the Sanity Challenge, Path Two: Vibe-Code Something Strange

What I Built

I built Vision Assistant v1.0.0, a custom desktop AI vision tool built from scratch using Electron. It is designed to bring instant, system-wide AI assistance directly to the desktop environment without the friction of constant context-switching and this version package is built only for windows user right now.

Key Features:

Global Shortcut Activation:
Runs in the background and triggers instantly from any active window or application using the Ctrl + Shift + Space shortcut.

Real-Time Chat & Vision:
Seamlessly process and analyze visual content using integrated AI capabilities right from the desktop.

Multiple Screen Capture:
Allows capturing and analyzing multiple screens or active display areas simultaneously, perfect for multi-monitor workflows.

Native Windows Experience:

Lightweight standalone Windows application packaged via electron-builder with a sleek dark/purple UI theme. )

Demo

Setup Guide & Walkthrough Video:
[https://drive.google.com/drive/folders/1Kci00sPzQalZA78Urge4v8mkXLHp4dbR?usp=sharing]

Explore the repository and release notes on GitHub:
GitHub Release Page

Grab the direct package:
Direct Download Link

You can download the latest Windows release bundle directly from the GitHub repository releases.

Code

GitHub Repository:

https://github.com/chhavigupta783/vission-assisstant-v1

My Build Process

AI-Native Setup:
Leveraged AI capabilities to vibe-code and structure the entire Electron desktop app from scratch, handling native window bindings and global shortcuts.

AI Integration:
Powered by the Gemini API to handle real-time chat, vision analysis, and contextual interactions seamlessly.

Challenges Faced:
Managing local environment configurations, ensuring smooth packaging with electron-builder, and handling window focus events for system-wide shortcuts.

Course-Correction:

Refined the build setup documentation to guide users properly through extracting zip assets and bypassing Windows SmartScreen warnings for custom unsigned executables.

Sanity Project Details

Note: While the core application is built natively using Electron and the Gemini API, Sanity's documentation was referenced for content modeling patterns and architecture guidelines.

[A sleek, professional 3D graphic featuring a vibrant purple glowing bubble that showcases the dark-mode UI of the Vision Assistant desktop application. At the bottom, a stylish UI element clearly highlights the core system-wide feature with the text "press ctrl + shift + space".]

(https://dev-to-uploads.s3.us-east-2.amazonaws.com/uploads/articles/0ja24tx9sglq2sco303q.png)

<!-- Thank you for visiting --!>

DEV Community: Chhavi Gupta

**Building Vision Assistant: A Custom Desktop AI Tool Built with Electron**

What I Built

Demo

Partner Technologies

Hackathon Experience

Say Hello to Vision Assistant: Your Instant, Shortcut-Triggered Desktop AI Tool

``This is a submission for the Sanity Challenge, Path Two: Vibe-Code Something Strange

What I Built

** Key Features: **

Lightweight standalone Windows application packaged via electron-builder with a sleek dark/purple UI theme. )

Demo

You can download the latest Windows release bundle directly from the GitHub repository releases.

Code

https://github.com/chhavigupta783/vission-assisstant-v1

My Build Process

Refined the build setup documentation to guide users properly through extracting zip assets and bypassing Windows SmartScreen warnings for custom unsigned executables.

Sanity Project Details

Note: While the core application is built natively using Electron and the Gemini API, Sanity's documentation was referenced for content modeling patterns and architecture guidelines.

Building Vision Assistant: A Custom Desktop AI Tool Built with Electron

Key Features: