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

Description and use

A 17‑agent AI swarm on a Pixel 7 generates platform‑specific social media posts from long text, using a quantized Phi‑2 model and llama.cpp for on‑device inference.

Problem

Running a multi‑agent AI system that generates platform‑specific social media posts entirely on a smartphone, balancing model size, performance, and battery life.

Outcome

The swarm successfully produces platform‑specific social media posts from long text on a smartphone, demonstrating efficient on‑device inference with a 2.7 GB quantized Phi‑2 model.

Workflow

• Load text input via Input Manager
• Summarize text with two independent Summarizer agents
• Extract keywords with two Keyword Extractor agents
• Analyze tone with two Tone Analyzer agents
• Profile each platform with four Platform Profiler agents
• Generate posts of short/medium/long length with four Post Generator agents
• Generate hashtags with two Hashtag Generator agents
• Aggregate and output posts

Evidence

I built a 17-agent AI swarm, running on my Pixel 7, that analyzes text and generates targeted social media content.

Source

I settled on Phi‑2, a 2.7 billion parameter model from Microsoft. It’s surprisingly capable for its size.

Source

I quantized the model down to Q4_K_M. This reduces the model size from ~5.5GB to around 2.7GB, making it manageable on my phone’s storage.

Source