yashab-cyber/opendroid

What it solves

OpenDroid provides a way to turn an Android phone into an autonomous agent capable of executing complex, multi-step tasks across various apps and system settings. Instead of just chatting, it can plan and perform actions like checking the weather, sending messages, and setting alarms automatically.

How it works

The system uses an autonomous agent engine to break down user commands into sequential steps with dependency tracking. It leverages a vision engine (via the Accessibility API) to analyze the screen and a multi-tier memory system (working, episodic, semantic, and procedural) to remember context and preferences. It integrates with 12 different LLM providers—including cloud-based options like Gemini and OpenAI, and local options like Ollama—with a smart fallback chain to ensure reliability.

Who it’s for

Android users who want a hands-free, AI-driven automation tool that can control their device's system settings, communication apps, productivity tools, and smart home devices.

Highlights

• Autonomous Execution: Plans, executes, verifies, and adapts multi-step tasks dynamically.
• Broad Device Control: Manages 60+ actions across system settings, communication, finance, and navigation.
• On-Device Model Support: Includes a manager for downloading and verifying local models (LiteRT-LM).
• Vision-Capable: Uses screen analysis to interact with apps that don't have standard APIs.
• Multi-Tier Memory: Maintains long-term facts and user-defined macro workflows.