Korner — Cross-Platform AI Desktop Assistant

Korner is a lightweight Agentic AI desktop assistant that allows users to capture screenshots and interact with AI instantly without switching between multiple applications or browser tabs. It provides seamless integration with an AMD Instinct MI300X GPU running GPT OSS 120B, enabling advanced text and visual analysis directly on your desktop.

Features

• Region Screenshot Capture Select a specific area of your screen for AI analysis using a transparent overlay and drag-to-select interface.

• AI Questioning with Images Combine screenshots and user text queries and send them to the GPT OSS 120B Vision API for instant answers.

• Floating Response Window View AI responses in a movable, resizable window with options to copy, pin, or close the response.

• Versatile Use Cases

  • Academic: Capture code snippets, charts, or lecture slides and receive explanations or summaries.
  • Work: Summarize meeting notes, generate agendas, or assist with project management.
  • Personal: Plan travel itineraries, organize health routines, or track learning progress.
  • Entertainment: Receive recommendations for events, exhibitions, and leisure activities.

• Flexible Interaction Modes Supports text, voice, and visual input. Users can type queries, record audio, or submit screenshots for analysis.

Platforms

• Windows 10/11

Build

• Windows 10/11
  Install Go
  Install Python
  Install Ollama

go install github.com/wailsapp/wails/v2/cmd/wails@latest
wails build --clean

Install(IMPORTANT)

• Windows 10/11

python -m pip install -U openai-whisper
ollama run qwen3-vl:4b
winget install --id Gyan.FFmpeg --source winget

Technical Overview

• Framework: Wails v2 (Go 1.22 + Vue 3)

• Frontend: Vue 3 + Tailwind CSS

• Backend: Go (handles screenshot capture, hotkeys, and LLM communication)

• LLM Integration: Connects to AMD GPT OSS 120B via OpenAI-compatible API

• Packaging:

  • Windows: .exe with system tray icon

Privacy and Security

• Screenshots are kept in memory only and are never written to disk.
• No user data is stored or shared without explicit consent.

Future Enhancements

Planned upgrades include:

• Personalized recommendations based on user activity
• Proactive alerts and reminders
• Advanced multi-modal reasoning for more intelligent assistant capabilities

Usage

1. Launch Korner and place the floating icon anywhere on your screen.
2. Use the global hotkey or click the icon to start a screenshot or AI interaction.
3. Submit your query (text, voice, or image).
4. View responses in the floating window and interact as needed.

Goal

Korner aims to provide a truly integrated AI assistant for desktop users. By reducing the need to switch between applications or browser-based LLMs, it enhances productivity, learning, and daily life efficiency across personal and professional tasks.

Demo

https://www.youtube.com/watch?v=28iCa3jkF10

Open-Source Credits

Thank you to the open-source community for making this tool possible!

Component	License	Links
FFmpeg	LGPL v2.1 / GPL v2+	https://ffmpeg.org · https://github.com/FFmpeg/FFmpeg
whisper	MIT	https://github.com/openai/whisper
Qwen3-VL series models	Tongyi Qianwen Research License (Apache-2.0 compatible)	https://huggingface.co/Qwen · https://github.com/QwenLM/Qwen3-VL
Ollama (for Qwen3-VL inference)	MIT	https://ollama.com · https://github.com/ollama/ollama