Glimmer: An Open-Source Android Context-Aware AI Companion System

Glimmer is an open-source, privacy-first Android context-aware AI companion system that breaks the boundaries of traditional isolated chat apps. It delivers a system-level digital companion experience via an interactive screen overlay. Key features include a local LLM engine, long-term memory system, modular plugin architecture, etc. It uses the GPL-3.0 license and emphasizes privacy-first design.

• Original Author/Maintainer: ALMLK1996
• Source Platform: GitHub
• Original Link: https://github.com/ALMLK1996/Glimmer-Assistant-
• Release Date: May 30, 2026

Unlike traditional AI tools that wait for commands, Glimmer coexists with daily digital workflows via an interactive screen overlay, learning user habits, managing tasks, and evolving emotionally. Its core philosophy is captured in the Arabic slogan: "ليس مساعداً ذكياً… بل كائن يعيش داخل هاتفك" ("Not just an intelligent assistant... but a being that lives inside your phone").

🎙️ Context Overlay Companion

A lightweight, non-intrusive screen overlay system that can be freely moved. It provides real-time context interaction, voice bubbles, and emotional feedback, allowing companionship without switching apps.

🧠 Local & Hybrid LLM Engine

Optimized for running Small Language Models (SLM) on the device, enabling 100% offline privacy protection. Optional encrypted hybrid cloud capabilities are available, and core data never leaves the device.

💾 Long-Term Memory System

Built-in secure local vector/relational storage that remembers conversations, adapts to user habits, and grows over time—key to a personalized experience.

🛡️ System Agent & Guardian Functions

A powerful automation engine orchestrates system tasks (opening apps, executing routines, handling notifications) while monitoring screen usage patterns to protect digital health.

🔌 Modular Plugin Architecture

Fully decentralized, allowing third-party developers to build custom skills, automation triggers, and extended features—with a continuously expanding ecosystem.

🎨 Open Skin Engine

Supports user-generated aesthetic layouts, custom themes, and full visual customization to create a unique digital companion appearance.

The project uses a clear layered architecture:

• core: Shared tools, dependency injection (Koin/Hilt), and theme foundations
• overlay: Overlay companion window management system (WindowManager)
• ai: LLM processing, local inference, and prompt engineering
• memory: Long-term local memory database (Room/encrypted SQLite)
• tasks: Background execution engine and OS automation agent
• plugins: Decoupled, extensible plugin interfaces
• guardian: Digital health, usage analyzer, and guardian mode

Additionally, it includes a complete documentation system (docs/), skin resource packs (skins/), and standardized open-source governance files (CONTRIBUTING.md, SECURITY.md, CODE_OF_CONDUCT.md).

Glimmer uses the GPL-3.0 open-source license, reflecting commitments to transparency and user freedom. The project emphasizes "privacy-first design" to ensure user data security through local processing and encrypted storage.

Glimmer represents a new paradigm for mobile AI interaction:

1. From Tool to Companion: An active partner in digital life, not a passive tool
2. Balance Between Privacy and Convenience: Local LLM inference balances privacy and intelligent services
3. Extensible Ecosystem: Plugin architecture supports community contributions, forming a virtuous cycle
4. Digital Health Guardian: Built-in usage monitoring helps establish healthy digital habits

For Android developers and AI enthusiasts, Glimmer is not only a usable product but also an excellent example for learning to build context-aware AI systems.

Glimmer demonstrates the possibility of AI assistants evolving from "in-app chat" to "system-level companions". Its open-source nature, privacy-first design, and modular architecture make it an emerging project worth watching in the mobile AI field.

Recommendations: Developers interested in mobile AI can refer to its technical practices; regular users can visit the GitHub repo to learn more and try it out; community members can participate in ecosystem building through plugin development or skin design.