EdgeBrain — A Complete Edge AI Platform That Runs 100% Locally (Open Source)

#showdev

What if you could run a full AI-powered smart home system without any cloud dependency?

That's exactly what EdgeBrain does. It's an open-source edge AI platform that simulates, processes, and acts on IoT sensor data — all locally, all free.

Why Edge AI Matters

Cloud AI has problems:

Latency — round-trips to the cloud add 200-500ms
Privacy — your smart home data sent to third parties
Cost — API calls add up fast
Offline — no internet = no intelligence

EdgeBrain solves all of these by running everything on your local machine.

What It Does

11 Simulated IoT Devices — temperature, motion, energy, humidity, light sensors across 3 rooms
Multi-Agent AI System — Data Agent → Decision Agent → Action Agent pipeline
3 Decision Strategies — Threshold Rules (with hysteresis), Anomaly Detection (Z-Score + IQR + Gradient), No-Motion Timeout
Predictive Analytics — Linear regression forecasting, moving averages, anomaly scoring
Real-Time Dashboard — 5-page React dashboard with live charts, device states, agent pipeline visualization
MQTT + WebSocket — Full IoT communication stack
Docker Compose — 6 services, one command to start everything

Quick Start

git clone https://github.com/rudra496/EdgeBrain.git
cd EdgeBrain
docker compose up --build -d
# Open http://localhost:3000

Architecture

The system uses a clean multi-agent pipeline:

Sensor → MQTT → Data Agent → Decision Agent → Action Agent → Alert/Command

Each agent has its own metrics, error handling, and performance tracking. The decision engine uses a plugin architecture — just implement the DecisionStrategy interface and register it.

Tech Stack

Layer	Tech
Backend	FastAPI + Python 3.11
Frontend	React + Recharts
Database	PostgreSQL 16
Cache	Redis 7.2
Messaging	Mosquitto MQTT
Infra	Docker Compose
Hardware	ESP32 (optional)