English

🤖 Welcome to BreezeApp Engine!

Ready to build the future of on-device AI on Android? You're in the right place!

BreezeApp Engine is a next-generation framework for creating powerful, modular, and extensible AI-driven experiences.

📦 What's in This Repository

This repository contains two main components:

1. BreezeApp Engine (android/breeze-app-engine)

The AI inference service that runs as a background Android service. It manages AI models, routes requests to appropriate runners, and handles the heavy lifting of AI processing.

Key Features:

• Multi-provider AI support (ExecuTorch, Sherpa, OpenRouter, LlamaStack)

• Model management and download

• Configuration UI for settings

2. EdgeAI SDK (android/EdgeAI)

The client-side SDK that provides an effortless, lightweight approach for apps to leverage AI features without embedding the entire engine. Apps simply bind to the engine service via AIDL and get a clean, type-safe API.

Key Features:

• Simple API: EdgeAI.chat(), EdgeAI.asr(), EdgeAI.tts()
• AIDL-based IPC for cross-process communication
• Streaming support for real-time responses

---

🔎 How the Engine Serves Requests

The engine accepts AIDL connections from client apps and routes requests to appropriate AI runners. Here's how the engine processes incoming requests:

%%{init: {'flowchart': {'useMaxWidth': false, 'width': 800}}}%%
graph TD
    subgraph Client["Client App Process"]
        A["📱 Client App<br/>(via EdgeAI SDK)"]
    end
    
    subgraph Engine["BreezeApp-engine Process"]
        B["🔌 AIDL Service Binder"]
        C["🎯 Request Router<br/>(AIEngineManager)"]
        D["🤖 AI Runners"]
        E["📊 Model Manager"]
    end

    A -- "1. bindService()" --> B
    B -- "2. Route request" --> C
    C -- "3. Select runner" --> D
    C -- "Load model if needed" --> E
    D -- "4. Return result" --> C
    C -- "5. Stream/return response" --> B
    B -- "6. Deliver to client" --> A

    style B fill:#FFE0B2,stroke:#FF6F00
    style C fill:#E1BEE7,stroke:#7B1FA2
    style D fill:#B2DFDB,stroke:#00796B
    style Client fill:#E8F5E9,stroke:#4CAF50,stroke-width:2px
    style Engine fill:#E3F2FD,stroke:#2196F3,stroke-width:2px

Loading

This architecture allows the engine to manage multiple AI providers and route requests efficiently.

🤖 Supported AI Providers

BreezeApp Engine integrates with multiple AI providers, each bringing unique capabilities:

Provider	Type	LLM	VLM	ASR	TTS	Guardian	Streaming
MediaTek	Local NPU	✅	❌	❌	❌	❌	✅
ExecuTorch	Local	✅	❌	❌	❌	❌	✅
LlamaStack	Remote	✅	✅	❌	❌	✅	✅
OpenRouter	Remote	✅	✅	❌	❌	❌	✅
Sherpa	Local	❌	❌	✅	✅	❌	✅

Legend: ✅ Supported | 🚧 Experimental | ❌ Not Supported

For detailed technical implementation, see Engine Architecture →

🚀 How to Get Started

Your path depends on your goal. Are you building an app with the engine, or building a new feature for the engine?

📱 For App Developers (Using the Engine)

If you want to add AI features to your Android app, this is your path.

1. Start Here: Our BreezeApp Client Guide is the best place to begin. It provides a step-by-step tutorial on how to integrate the EdgeAI SDK and make your first API call.
2. Explore the API: The EdgeAI SDK is the public API for the engine. You can explore its features and data models in the EdgeAI README.

🧠 For AI/ML Engineers (Extending the Engine)

If you want to add a new model or AI capability to the engine itself.

1. Understand the Design: The Architecture Guide explains the internal design of the engine and how all the pieces fit together.
2. Build a Runner: Follow the Runner Development Guide to learn how to implement a new Runner that can be discovered by the engine.

⚠️ Development Note: After rebuilding the engine, you must also rebuild your client app to reconnect properly.

🤝 Join Our Community & Contribute

Whether you're fixing a bug, improving the docs, or adding a revolutionary new runner, we welcome your contributions!

• Contribution Guidelines: Please read our Contributing Guide.
• Have Questions? Don't hesitate to open an issue! We're happy to help.