_{Logo designed by Sushant Bhattarai}

Debloat Android. De-bloat. Destroy bloatware.
A local-first desktop app to safely remove Android bloatware on non-rooted devices.

De-Droid

De-Droid combines local ADB execution with safety intelligence to help you remove unnecessary apps with confidence.

• Local ADB actions for uninstall, disable, enable, and restore.
• Safety labels (RECOMMENDED, ADVANCED, EXPERT, UNSAFE) with confidence signals.
• Smart filtering for system/user, bloatware, state, and removal level.
• Backup snapshots, history tracking, and before/after comparisons.
• Permission and background control insights per package.
• App Store for some most have to open source apps from Fdroid.
• Optional ML-assisted package safety scoring via FastAPI.

Architecture

De-Droid follows a local-first architecture:

• Desktop app (desktop-app/): Electron + React + TypeScript UI and local command execution.
• Model API (model-api/): optional FastAPI service for advanced safety predictions.
• Local data: JSON + SQLite for package metadata, backups, history, telemetry, and settings.

See ARCHITECTURE.md for full architecture details.

Repository Structure

de-droid/
  desktop-app/          # Electron + React + TypeScript app
  model-api/            # ML pipeline + FastAPI service
  LICENSE               # MIT license
  README.md

Quick Start

Requirements

• Node.js 20+
• npm
• Python 3.11+ (only for optional model API)

Note: De-Droid supports bundled platform-tools for packaged releases, so end users do not need to install ADB manually.

Run Desktop App

cd desktop-app
npm install
npm run dev

Run Model API (Optional)

python -m venv model-api/env
source model-api/env/bin/activate
pip install -r model-api/requirements.txt
uvicorn main:app --app-dir model-api/model-api --host 0.0.0.0 --port 8000 --reload

Train and Export Predictions (Optional)

python model-api/scripts/build_training_dataset.py
python model-api/scripts/train_safety_model.py --predictions-out-desktop desktop-app/src/data/safety_predictions.json

Safety Labels

• RECOMMENDED: usually safe for most users.
• ADVANCED: needs review based on your use case.
• EXPERT: may affect features; remove only if you understand impact.
• UNSAFE: removal is blocked/restricted for safety.

Model API Endpoints

• GET /health
• POST /api/check-packages
• POST /api/check-single
• GET /api/stats
• GET /api/critical-packages
• GET /api/search/{query}

License

This project is licensed under the MIT License. See LICENSE.