Dispatch Alpha

Dispatch Alpha is an intelligent radio dispatch system that integrates Baofeng walkie-talkies with a Raspberry Pi and AI capabilities. Using OpenAI's Whisper API for transcription, ChatGPT for generating responses, and Text-to-Speech (TTS0-1) for audio output, the system can transcribe radio communications, analyze them, and respond to user inputs on the same radio channel.

Watch the demo

Overview

Dispatch Alpha aims add a dispatcher to any channel by introducing AI capabilities from OpenAI to traditional radio systems. Currently using two baofengs and a raspberry pi. You also need an additional baofeng for just listeing in. I'm using these and they're great:

Baofeng Radios $70 for two radios

USB Sound Card $7

2.5mm to 3.5mm audio cable $6

• Receiver (RX): Listens to the radio channel, captures incoming audio, and transcribes it using OpenAI's Whisper API.
• Transmitter (TX): Generates AI-based responses using OpenAI's ChatGPT and TTS APIs, and transmits them back over the radio channel.

This is what my setup looks like:

    │
[Raspberry Pi]
    │
    - [USB Sound Card] -> [RX Baofeng (Receiver)]
    │ - 2.5mm audio cable (Speaker Output) -> Mic-In (Red Port)
    │ (RX code listens here src/reciever/rx-stable.py) 
    │
    - [ BCM2835 Headphone Jack (speaker output) ] -> [TX Baofeng (Transmitter)]
    │ - 3.5mm audio cable -> Mic Input of Baofeng
    │ (TX code listens here src/transmitter/tx-stable.py)
    │

Features

• Real-Time Transcription: Captures and transcribes incoming radio communications.
• AI-Powered Responses: Generates context-aware responses using OpenAI's ChatGPT.
• Text-to-Speech Conversion: Converts AI-generated text responses to speech for radio transmission.
• Conversation History: Maintains a conversation history to provide context-aware interactions.
• Configurable Settings: Adjustable thresholds and parameters for audio processing.
• Logging: Comprehensive logging of transcriptions and system operations.

Hardware Setup

Components

• Raspberry Pi (using an rpi4 with 1G RAM and 128 SD disk)
• Two Baofeng Walkie-Talkies
• USB Sound Card
• Audio Cables and Adapters

Connections

Receiver (RX) Baofeng:

• Speaker Output (2.5mm jack): Connects to the Mic-In (red port) on the USB sound card using a 2.5mm male-to-male cable or appropriate adapter.
• PTT (Push-To-Talk): Not required for RX.

Transmitter (TX) Baofeng:

• Mic Input (3.5mm jack): Connects to the Raspberry Pi's audio output (headphone jack or USB sound card output).
• PTT Control: Configured to use VOX mode for automatic transmission upon detecting audio output from the Raspberry Pi.

USB Sound Card:

• Plugs into one of the Raspberry Pi's USB ports.
• Provides mic-in and audio-out capabilities.

Connection Diagram

[RX Baofeng Speaker Output] --(2.5mm audio cable)--> [USB Sound Card Mic-In (Red Port)] --(USB)--> [Raspberry Pi]

[Raspberry Pi Audio Output] --(3.5mm audio cable)--> [TX Baofeng Mic Input (3.5mm jack)]

Software Setup

Requirements

• Raspberry Pi with Raspbian OS (or similar Debian-based OS)
• Python 3.7 or higher
• OpenAI API Access (for Whisper, ChatGPT, and TTS APIs)
• Python Libraries:
  • numpy
  • sounddevice
  • soundfile
  • openai
  • asyncio
  • threading
  • logging

Configuration

1. OpenAI API Key: Obtain an API key from OpenAI and set it as an environment variable:

   export OPENAI_API_KEY='your_actual_api_key_here'

2. Audio Device Index: Adjust the AUDIO_DEVICE_INDEX in both rx-stable.py and tx-stable.py to match your audio devices. You can list audio devices using the list_audio_devices() function or by running:

   python3 -c "import sounddevice as sd; print(sd.query_devices())"

3. Adjust Thresholds: Modify the AUDIO_THRESHOLD, SILENCE_DURATION_THRESHOLD, and other constants in the scripts to fine-tune audio processing based on your environment.

Installation

1. Clone the Repository

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

2. Set Up the Environment Ensure your Raspberry Pi is up to date:

   sudo apt-get update
   sudo apt-get upgrade

3. Install Python libraries:

python -m venv .venv
source .venv/bin/activate
pip3 install -r requirements.txt

4. Configure Audio Devices List available audio devices:

   python3 -c "import sounddevice as sd; print(sd.query_devices())"

   Identify the indexes of your input and output audio devices and update config/settings.py:

   AUDIO_DEVICE_INDEX = 2  # This is for RX
   TX_AUDIO_DEVICE_INDEX = 1  # This is for TX
   AUDIO_DEVICE = "plughw:1,0"  # ALSA name for the onboard headphone jack

5. Set Up OpenAI API Key

   Set your OpenAI API key as an environment variable:

      export OPENAI_API_KEY='your_actual_api_key_here'

Receiver (RX) Script

The rx-stable.py script captures audio from the RX Baofeng, sends it to OpenAI's Whisper API for transcription, and saves the transcribed text.

Running the RX Script:

bash scripts/run_receiver.sh

Command-Line Options:

• --test: Run in test mode, records for 5 seconds, and exits.
• --debug: Enable debug mode to save audio files to disk.

Functionality:

• Audio Capture: Captures audio from the specified input device.
• Voice Activity Detection: Detects when someone is speaking based on audio levels.
• Transcription: Sends audio chunks to OpenAI's Whisper API for transcription.
• Logging: Logs transcriptions and saves them to the transcriptions directory in JSON format.

Transmitter (TX) Script

NOTES ON TX:

• Configure VOX mode on the Baofeng. Sensitivity is set between 0 (off) and 10. With 1 being the most sensitive. Choose 1.
• Using alsamixer set the volume of the on-board headphone jack to 90+ to trigger the VOX.

The tx-stable.py script monitors transcriptions, generates responses using OpenAI's ChatGPT, converts the responses to speech using OpenAI's TTS API, and plays the audio back through the TX Baofeng.

Running the TX Script:

bash scripts/run_transmitter.sh

Command-Line Options:

• --debug: Enable debug mode to save TTS audio files to disk.

Functionality:

• Monitoring Transcriptions: Watches the transcriptions directory for new transcriptions.
• AI Response Generation: Uses OpenAI's ChatGPT to generate context-aware responses when addressed.
• Text-to-Speech Conversion: Converts AI-generated text responses to speech using OpenAI's TTS API.
• Audio Playback: Plays the audio responses through the specified output device.

Note: The script assumes that the TX Baofeng is set to VOX mode to automatically transmit when audio is played.

Example Interaction:

1. User: "Dispatch Alpha, we have a situation at the main gate."
2. System: (Transcribes the message and detects that it addresses the dispatcher.)
3. System: Generates a response: "Copy that. Dispatch out, 14:35."
4. System: Converts the response to speech and transmits it over the radio.

Future Development

Thinking about some additions..

• Enhanced AI Capabilities: Analyze transcriptions for commands, keywords, or emergency signals.
• Web Dashboard: Develop a web-based interface for monitoring and controlling the system.
• Custom Voice Models: Integrate custom TTS voices for more personalized interactions.

Contributing

Contributions are welcome! Please open an issue or submit a pull request on GitHub.

License

MIT License