MiQ

MiQ, short for Machine Instruction Query, is a desktop assistant that combines a Java Swing client with a FastAPI backend. The project focuses on voice-driven interaction, intent classification, shortcut management, and local AI-assisted responses.

Highlights

• Voice-triggered desktop assistant flow
• Intent detection for actions such as open, add, remove, list, chat, and exit
• User shortcut management backed by a database
• Local chat integration support for private/offline experimentation
• Java desktop UI paired with a Python AI/API backend

Architecture

• frontend/: Java 21 Swing desktop app
• backEnd/: FastAPI backend, database access, and AI integration

Tech Stack

• Frontend: Java 21, Maven, Vosk, JNativeHook
• Backend: Python, FastAPI, SQLAlchemy, Transformers, RapidFuzz
• Local AI tooling: Hugging Face Transformers, optional GGUF chat model flow

Running Locally

Backend

1. Create and activate a Python virtual environment.
2. Install dependencies:

cd backEnd
pip install -r ai/requirements.txt

3. Start the API server:

uvicorn main:app --reload

If you have a locally exported intent model, point the app to it with MIQ_MODEL_DIR. Chat GGUF assets can be downloaded locally with the helper in backEnd/ai/download_chat_model.py.

Frontend

1. Install Java 21 and Maven.
2. Add any local runtime assets you want to use under frontend/lib/ such as Vosk models and optional jars.
3. Start the desktop app:

cd frontend
powershell -ExecutionPolicy Bypass -File .\run-dev.ps1

The included icon and activation sound now load from the project directory instead of a machine-specific path.

Repo Notes

• Dataset files are intentionally excluded from version control.
• Large trained model folders, generated output, IDE files, and caches are ignored.
• Some runtime assets are expected to be restored locally instead of being committed.