drwjkirkpatrick-web/poultry-guardian
Domain:
agricultureRecord type:
software
Creator:
drw
Host:
AI-powered autonomous poultry monitoring system for smallholder African farmers. Raspberry Pi 4/5, under USD150/coop, 2G GSM connectivity.
# HERMES Poultry Guardian
**AI-Powered Autonomous Poultry Monitoring System for Smallholder African Farmers**
| | |
|---|---|
| **Hardware** | Raspberry Pi 4 / 5 (4 GB ARM64) |
| **AI Model** | MiniMax 2.7 (via Ollama Cloud) |
| **Language** | Python 3.10+ |
| **Cost** | Under USD150 per coop |
| **Connectivity** | 2G GSM OK (no broadband required) |
---
## What It Does
HERMES Poultry Guardian monitors temperature, humidity, ammonia (NH₃), water level, feed level, and chick behavior via USB webcam — then automatically controls fans, heaters, and window actuators to keep poultry safe. It sends SMS/WhatsApp alerts via Twilio with SIM800L GSM fallback for remote areas.
**Five AI agents:**
- **Vision Agent** — YOLOv8n chick behavior detection (clustering, heat stress, inactivity, mortality)
- **Prediction Agent** — Open-Meteo weather forecasting + temperature trend spike detection
- **Supervisor Agent** — Priority-based decision orchestration with 5-tier graduated heating/cooling and ammonia-compensated ventilation
- **Cleaning Agent** — Coop hygiene tracking with NH₃ alert gating (6-hour post-cleaning grace, recontamination detection)
- **Learning Agent** — Adaptive fan speed via exponential moving averages (EMA) per speed tier
---
## Quick Start
```bash
# Clone
git clone github.com
cd poultry-guardian
# Install dependencies
python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt
# Configure
cp config/.env.example config/.env
# Edit config/.env with your Twilio credentials and phone numbers
# Simulate (no hardware needed)
POULTRY_SIMULATE=true python3 src/main.py --test-agents
# Deploy for real
python3 src/main.py
```
---
## Architecture
```
src/
├── main.py # Orchestrator — loop, agent coordination, shutdown
├── agents/
│ ├── supervisor_agent.py # Priority decision tree + graduated control
│ ├── vision_agent.py # OpenCV + YOLOv8 behavior detection
│ ├── …