About Dispatch460

Since 15-Feb-2026 until 11-may-2026 5:05pm

đŸ“ģ What is Dispatch460?

Dispatch460 is a real-time radio monitoring system that captures, transcribes and analyzes emergency dispatch communications from Milford, CT Fire and EMS on frequency 154.340 MHz. Every transmission is automatically recorded, converted to text using AI, geocoded to a map location, and made available through this web interface within seconds of broadcast.

The system is completely automated — a small computer listens to the radio 24 hours a day, 7 days a week, and pushes every transmission to the cloud in real time. Community members can follow live emergency activity, review historical dispatches, and explore patterns in emergency response across Milford.

🌱 How It Started

Dispatch460 began as a simple experiment — a Raspberry Pi-style mini-PC connected to an inexpensive software defined radio (SDR) dongle, listening to the Milford Fire/EMS frequency. Early versions simply recorded audio clips and stored them locally.

Over time the system grew more sophisticated. OpenAI's Whisper speech recognition model was integrated to automatically transcribe each transmission. A custom fine-tuned version of Whisper was trained specifically on Milford radio traffic to improve accuracy on local street names and dispatch terminology. The system then migrated from local Flask web servers to a fully hosted Django application on AWS, making it accessible from anywhere in the world.

📈 Continuous Improvement

Dispatch460 is designed to get better every day through an active learning pipeline:

Daily — Data Capture

Every transmission is captured, transcribed and geocoded automatically. A street name correction dictionary fixes known Whisper transcription errors in real time.

Nightly — AI Quality Review

An AI agent powered by Claude analyzes the day's dispatches, flags geocoding failures, transcription errors and classification issues, and suggests corrections.

Weekly — Data Quality Report

A natural language data quality report is automatically generated, scoring the day's data and listing specific issues to investigate.

Monthly — Model Retraining

Approved corrections are exported as training pairs and used to fine-tune the Whisper transcription model, making it progressively more accurate on Milford-specific terminology.

đŸ› ī¸ Technology Stack

RTL-SDR DongleRadio hardware
OpenAI WhisperAI transcription
DjangoWeb framework
PostgreSQLDatabase (AWS RDS)
Amazon S3Audio storage
Elastic BeanstalkCloud hosting
Google Maps APIGeocoding
Leaflet.jsInteractive maps
Claude APIAI quality analysis
DragonOS LinuxCapture platform
Python 3.11Core language
No-IP DDNSDynamic DNS

📊 System Statistics

154.340
MHz monitored
24/7
Continuous monitoring
17388
Transmissions captured
1736
Dispatches geocoded

âš ī¸ Disclaimer

Dispatch460 is an independent community project and is not affiliated with, endorsed by, or operated by the City of Milford, Milford Fire Department, or Milford EMS. All radio transmissions monitored are publicly broadcast on licensed frequencies. Transcriptions are generated by AI and may contain errors. This system is intended for informational purposes only and should never be used as a substitute for official emergency communications.