Core Dispatch – AI-Driven Radio Communication System

Core Dispatch

Core Dispatch is an AI-powered radio communication system that transforms classic handheld two-way radios into an intelligent, context-aware presence on the airwaves. By combining Raspberry Pi, OpenAI’s GPT-based models, and Text-to-Speech (TTS), Core Dispatch can receive, transcribe, process, and respond to radio communications in real time—whether for emergency dispatch, warehouse inventory updates, or just a laid-back “The Dude” persona for fun.

Overview

Core Dispatch injects AI-driven personas into any channel you choose, using three handheld two-way radios (one for receiving, one for transmitting, and one for the operator), plus a Raspberry Pi. It listens for transmissions, transcribes them (e.g., via OpenAI Whisper), and generates persona-specific responses (e.g., a chill “Dude,” a knowledgeable “Warehouse Worker,” or a “Dispatch” lead for emergencies). It then uses TTS to play those responses back over the same channel.

Key Components

• Receiver (RX): Listens for audio above a certain threshold, transcribes it with Whisper or another STT service.
• Transmitter (TX): Takes AI-generated text, converts it to speech using TTS, and broadcasts via VOX.
• Multiple Personas: Easily swap between “Dispatch,” “Fire Analyst,” “Warehouse Worker,” or a custom persona for your scenario.

Hardware Setup

Below is the minimal hardware list for Core Dispatch:

1. Raspberry Pi (4, 5, or similar)
2. Three handheld two-way radios (one for RX, one for TX, one for the operator to talk from)
3. USB Sound Card (for capturing audio from RX radio)
4. 2.5mm → 3.5mm audio cable (RX radio speaker out to Pi mic in)
5. 3.5mm → 3.5mm cable (Pi headphone out to TX radio mic in)
6. VOX on TX radio for automatic push-to-talk

Approximate cost is around $200 on Amazon, depending on deals.

Wiring Diagram

[RX Radio] --(Speaker Out)--> [USB Sound Card Mic-In] --(USB)--> [Raspberry Pi]
[Raspberry Pi Headphone Out] --(3.5mm cable)--> [TX Radio Mic-In]

1. Set VOX on your transmit radio so it keys up when the Pi plays audio.
2. Check device indexes on your Pi with sounddevice.query_devices() and update the config accordingly.

FCC Regulations & Frequency Use

If you’re in the United States:

Unlicensed Operation: Family Radio Service (FRS) channels typically do not require an FCC license, but they do have power/output limitations. GMRS (General Mobile Radio Service) requires a license, but offers higher power allowances. Amateur (Ham) Bands: You must have at least a Technician license (or higher) to transmit, and you must identify your callsign according to Part 97 rules. Always Verify Local Rules: Regulations vary by region or country (e.g., PMR446 in Europe).

Disclaimer: This project does not provide legal advice. Use only frequencies you’re licensed (or license-free) to operate on.

Software Setup

1. Clone the Project

   git clone https://github.com/jmcdice/core-dispatch.git
   cd core-dispatch
   

2. Install Dependencies

   python -m venv .venv
   source .venv/bin/activate
   pip install -e .
   

3. Set Environment Variables

   export OPENAI_API_KEY="sk-xxxx"
   export TRANSCRIPTION_SERVICE_TYPE="openai-whisper"
   export TTS_PROVIDER="openai"
   

4. Launch the Receiver

   python start_reciever.py
   

   • Continuously listens for radio traffic, transcribes speech, and saves JSON transcription files.
5. Launch the Transmitter

   python start_transmitter.py --profile emg_response
   

   • Monitors transcriptions and responds via Dispatch, Fire Analyst, or Logistics personas.

Features

• Real-Time Transcription: Captures incoming radio transmissions and converts them to text (Whisper, Google Chirp, or a mock STT).
• AI-Powered Personas: Load any persona (or multiple personas) with custom system prompts, specialized behavior, and activation phrases.
• Tool Integration: GPT-4 can call external “tools” (e.g., inventory lookups) to fetch real data for responses.
• TTS Playback: Multiple TTS engines supported (OpenAI, UnrealSpeech). The final speech is piped through the TX radio on VOX.
• Context-Aware Conversation: Maintains conversation logs so each persona can provide relevant, context-specific replies.
• Logging: Comprehensive logs for debugging and reviewing activity.

Example Scenarios

1. Emergency Dispatch
   • “Dispatch, this is Unit 33—fast-moving fire near Sunset Blvd.”
   • Fire Analyst and Logistics personas can step in with spread rates and water tanker availability.
2. Warehouse Worker
   • “Warehouse, do we have organic almond milk in stock?”
   • The AI persona calls an InventoryLookupTool and replies with item quantity and aisle number.
3. Just Having Fun
   • “Hey Dude, how about a round of bowling?”
   • The Dude persona responds in a chill manner, referencing White Russians and abiding by the lanes.

Potential Use Cases

• Incident Management: Fire/EMS dispatch, situational awareness, quick queries and resource checks.
• Business Applications: Warehouse or retail inventory queries, stock notifications, or shipping updates.
• Novelty or Entertainment: A “DJ” persona that plays music clips or a “gamer buddy” persona for on-air banter.

Next Steps

• Add More Personas: Build a “Security Bot,” “Translator Bot,” or “DJ Dave” persona.
• Try Different TTS/STT: Swap Whisper for Google Chirp, or OpenAI TTS for UnrealSpeech.
• Custom Tools: Link your database or API so the AI can answer real-time queries.
• Hardware Upgrades: Experiment with better microphones, external speakers, or even multiple Pi setups networked together.

Closing Thoughts

Core Dispatch bridges the gap between traditional radio communication and cutting-edge AI. Whether you’re managing emergency scenarios, running a warehouse, or just having fun with an AI “buddy” over the air, the system’s modular design makes it easy to innovate.

Stay tuned for future updates, and feel free to fork and contribute on GitHub. Got ideas or questions? Let us know!

---

Thanks for reading! If you found this project exciting, share it with your fellow radio enthusiasts, AI buffs, or open-source friends. Happy (AI) radioing!