WidgetAI

WidgetAI is a floating desktop widget that lets users send prompts to ChatGPT, Claude, or Perplexity and read the responses directly inside a compact Apple-style interface.

It is designed to remove browser friction: no tab switching, no manual copy-pasting, and no need to manage multiple AI websites separately.

---

Overview

WidgetAI starts as a compact, wide desktop widget inspired by Apple’s medium widget layout, then expands into a larger panel when a response is available. The visible interface stays minimal, while the actual AI interaction runs through a hidden embedded browser powered by Qt WebEngine.

The result is a lightweight desktop workflow for querying multiple AI providers from one persistent floating widget.

---

Features

• Floating always-on-top widget: stays visible above other windows for fast access while working.
• Compact-to-expanded layout: rests in a medium-style format and expands automatically when a response arrives.
• Unified AI access: supports ChatGPT, Claude, and Perplexity through one consistent command interface.
• Inline response display: answers are shown directly inside the widget in a read-only output panel.
• Background browser automation: prompts are sent through a hidden Qt WebEngine worker window, keeping the user-facing UI clean.
• Command history support: recent prompts can be viewed from inside the widget.

---

Pre-requisites

pip install -r requirements.txt

Docker

Docker image is available, users can pull and run WidgetAI directly from Docker Hub.

Pull the image

docker pull pa9da/widgetai:latest

---

Usage

Enter a slash command followed by a prompt, then press Enter.

/chatgpt explain how sorting algorithms work
/claude write a cover letter for a software engineering role
/perplexity what is the latest news about the James Webb telescope

---

Note

The project is under construction, more features yet to come.

TODO:

1. Login issue when initially using ai.
2. Proper message output for tables and images.
3. History chat per model and use history function as memory if model usage goes down append history to other chatbots.
4. Model usage limit.
5. Command teriminal / keyboard command / deployment