🔊 Acoustic Guard — Infrastructure Hub

Complete Docker Compose infrastructure for the Acoustic Guard distributed urban acoustic monitoring system — provisioning core backend services, ML classifiers, messaging brokers, and edge agents for both simulation and live testing.

---

📐 Architecture Overview

The infrastructure is divided into two logical blocks: Core Services and the Edge Network.

Core Services

Service	Stack	Role
rabbitmq	RabbitMQ 3	AMQP message broker for high-throughput audio event streams
postgres	PostgreSQL 15 + PostGIS	Spatial database for alerts, incidents, sensors, telemetry
classifier	Python + YAMNet + Random Forest	gRPC server with dual ML pipeline for threat classification
java-hub	Java + Spring Boot	Central orchestrator — REST API, WebSocket broadcaster, DB persistence

Classified threat types: UAV · Explosion · Siren · Truck · Generator

---

🚀 Modes of Operation

The system supports two distinct ways to run the infrastructure.

---

🎮 Mode 1 — Full Simulation (default, out-of-the-box)

Spins up the entire backend along with 13 virtual edge agents streaming pre-recorded anomalies (explosions, sirens, UAVs) across different geographic zones.

1. Start the cluster

docker compose up -d --build

2. Open the frontend

http://localhost:5173

3. Log in with the default admin credentials

Username: admin
Password: admin123

4. Observe — the dashboard populates with live simulated incidents, telemetry, and the noise heatmap in real time.

---

🎙️ Mode 2 — Live Microphone Testing

Use this mode to test the ML classifier and backend using a physical microphone through a local Rust edge agent.

Step 1 — Windows Audio Setup (crucial for accurate classification)

To prevent OS-level audio processing from distorting ML input:

1. Open Sound Control Panel → Recording → Microphone Properties
2. Enhancements tab → check "Disable all sound effects" (disable any AI noise cancellation/suppression)
3. Advanced tab → set Default Format to a clean sample rate without spatial audio

Step 2 — Start Core Infrastructure

This starts the Database, RabbitMQ, Classifier, and Hub — without the 13 simulated agents, so your microphone is the only data source:

docker compose -f docker-compose.live.yaml up -d --build

Step 3 — Start the Local Edge Agent

cd ../edge-agent
cargo run

Ensure the edge agent is configured to use AUDIO_SOURCE=MIC rather than FILE mode.

Step 4 — Test in Real Time

Log into the frontend (admin / admin123), play a siren or generator sound from your phone near the microphone, and watch the system detect it live.

---

🛠 Prerequisites

Requirement	Version	Notes
Docker	v24.0.0+	Required for all modes
Docker Compose	v2.20.0+	Required for all modes
Rust / Cargo	latest stable	Mode 2 only
Free RAM	4–8 GB	17 containers run concurrently in Mode 1

---

🔌 Exposed Ports

Service	Port	Description	Credentials
Java Hub (REST + WS)	8080	Main API & WebSocket endpoint	—
PostgreSQL	5432	Direct database access	postgres / password
RabbitMQ AMQP	5672	AMQP port for edge agents	guest / guest
RabbitMQ UI	15672	Management web interface	guest / guest
gRPC Classifier	50051	Direct ML inference endpoint	—

---

🧹 Maintenance & Cleanup

Stop all services gracefully:

docker compose down

Stop and wipe all data (destroys persistent volumes — database, logs, RabbitMQ queues):

docker compose down -v