Deepgram Flux Composite Agent

A simple terminal-based Python application that demonstrates the Deepgram Flux API integrated with OpenAI and Deepgram TTS. This demo processes a static audio file and creates a basic agent experience entirely in the terminal.

Features:

• 🎤 Flux Transcription: Real-time transcription using Deepgram's Flux API
• 🤖 OpenAI LLM: Processes audio to respond like an agent
• 🔊 Deepgram TTS: Natural voice synthesis for agent responses
• 🖥️ Terminal-only: No UI needed - everything runs in your terminal
• 📁 Static Audio: Processes your converted linear16 audio files

What It Does

The application follows this simple pipeline:

1. 📁 Audio Loading: Reads your converted linear16 audio file
2. 🎤 Flux Transcription: Sends audio to Deepgram Flux API for real-time transcription
3. 🤖 OpenAI Response: Generates intelligent responses using GPT-4o-mini
4. 🔊 TTS Generation: Converts the response to speech using Deepgram TTS
5. 💾 Audio Output: Saves the agent's speech as audio/responses/agent_response.wav

Getting an API Key

🔑 To access the Deepgram API you will need a free Deepgram API Key.

🔑 To access the OpenAI API you will need a free OpenAI API Key.

Installation

1. Install the required packages:

   pip install deepgram-sdk python-dotenv

Running the Demo

1. Create a .env file with your API keys:

   DEEPGRAM_API_KEY=your_deepgram_api_key_here
   OPENAI_API_KEY=your_openai_api_key_here

2. Use the default audio file provided OR add your audio file to the audio/ directory and convert to linear16 format:

   # Convert your audio file to the required format for Flux
   ffmpeg -i audio/your_file.wav -ar 16000 -ac 1 -c:a pcm_s16le audio/your_file_linear16.wav

   Then update the AUDIO_FILE path in main.py to point to your converted file.

3. Run the demo:

   python main.py

Example Output

Customization

You can modify the following in main.py:

• TTS model: Change "aura-2-phoebe-en" to other voices (aura-2-apollo-en, etc.)
• Audio filename: Change AUDIO_FILE to use a different filename
• Response logic: Modify the response generation to create different agent behaviors

Troubleshooting

1. Missing API Key: Set DEEPGRAM_API_KEY and OPENAI_API_KEY environment variables
2. Missing Audio File: Add audio file to audio/ directory and convert with FFMPEG
3. Wrong Audio Format: Flux requires linear16 - use FFMPEG command above
4. SDK Import Error: Install the Deepgram SDK correctly
5. Connection Issues: Check internet connection and API key validity

Documentation

You can learn more about Deepgram APIs at developers.deepgram.com.

Getting Help

We love to hear from you! If you have questions:

• Join the Deepgram Github Discussions Community
• Join the Deepgram Discord Community