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

Description and use

The author built a 17-agent AI swarm on a smartphone using Python, PyTorch Mobile, Kivy, and SQLite, employing quantization, pruning, and multithreading to operate within limited resources.

Problem

Run a complex, distributed AI system entirely on a smartphone, achieving emergent behavior while respecting mobile constraints such as limited CPU, RAM, and battery.

Outcome

The author successfully deployed a 17-agent AI swarm on a smartphone, demonstrating that distributed on-device AI is feasible with careful model optimization and system design.

Workflow

• Define agent roles and responsibilities
• Convert DistilBERT models to PyTorch Mobile format
• Quantize and prune models to 8-bit
• Implement agents in Python with Kivy UI
• Store agent memories in SQLite
• Set up a central message queue with tags
• Run agents in separate threads for concurrency
• Cache tokenization results
• Batch processing where memory allows
• Monitor memory and battery usage
• Iterate on optimization techniques

Evidence

Running 17 AI models on a phone requires serious optimization.

Source

PyTorch Mobile allows you to quantize your models to 8-bit integers, significantly reducing their size and improving inference speed.

Source

The agents don’t directly call each other’s methods. Instead, they communicate through a central message queue (implemented with Python lists and locks).

Source