I built a 17-agent AI swarm on my phone — here's how

说明与用途

在Android手机上完全离线运行的17代理AI群集，利用WebAssembly提升性能，Flutter访问传感器，实现了自发的活动和手势识别。

问题

Run a multi-agent AI swarm entirely on an Android phone without cloud APIs, achieving real-time interaction, privacy, and low latency.

结果

The swarm processes sensor data in real-time on-device, achieving activity and gesture recognition without cloud APIs, demonstrating emergent behaviour from simple agents.

工作流

• 1. Choose TypeScript for type safety and native web worker execution.
• 2. Use vanilla JavaScript with gl-matrix for vector math.
• 3. Compile performance-critical C++ code (smoothing, pattern recognition) to WebAssembly via Emscripten.
• 4. Load WASM modules into the JavaScript environment.
• 5. Build a Flutter Android app to host a headless webview, access sensors, and display UI.
• 6. Implement a simple in-memory message broker for agent communication.
• 7. Create 17 agents: 3 sensor input, 4 pattern recognition, 3 spatial mapping, 2 navigation, 3 action selection, 2 event logger.
• 8. Each agent processes data, publishes/subscribes via the broker.
• 9. Run the swarm, observe emergent behaviour (activity, gesture, environmental awareness).

证据

I built a 17-agent AI swarm that runs entirely on my Android phone. Not through cloud APIs (mostly), but actual on-device processing.

来源

The biggest bottleneck was processing the sensor data. JavaScript, while improved, isn't ideal for number-crunching. That's where WASM came in.

来源

This made a massive difference. The WASM-compiled code runs several times faster than the equivalent JavaScript, allowing the swarm to process sensor data in real-time without draining the battery.

来源

The swarm can reliably detect when I’m walking, running, or sitting still, based solely on the sensor data.

来源