DeepSeek R1 AI Voice Agent

A real-time AI voice assistant powered by DeepSeek R1 that enables seamless voice conversations through speech-to-text transcription, AI response generation, and text-to-speech synthesis.

🌟 Overview

This project creates an interactive AI voice agent that:

• Captures and transcribes speech in real-time using AssemblyAI
• Generates intelligent responses using DeepSeek R1 (7B model) via Ollama
• Converts AI responses back to natural speech using ElevenLabs
• Streams audio responses for immediate playback

✨ Features

• Real-time Speech Recognition: High-quality speech-to-text transcription with AssemblyAI
• Advanced AI Responses: Powered by DeepSeek R1's reasoning capabilities
• Natural Voice Synthesis: Professional text-to-speech with ElevenLabs
• Streaming Audio Playback: Low-latency audio streaming for responsive conversations
• Conversation Memory: Maintains context throughout the conversation
• Cross-platform Support: Works on macOS, Linux, and Windows

🔧 Prerequisites

API Keys Required

• AssemblyAI API Key: Get your free API key
• ElevenLabs API Key: Sign up for ElevenLabs

System Dependencies

Install Ollama

Download and install Ollama from ollama.com

Install PortAudio

Ubuntu/Debian:

sudo apt update && sudo apt install portaudio19-dev

macOS:

brew install portaudio

Windows: PortAudio is typically included with the Python package installation.

Install MPV (macOS only)

brew install mpv

📦 Installation

1. Clone the Repository

git clone https://github.com/danieladdisonorg/DeepSeek-R1-Voice-Agent.git
cd DeepSeek-R1-Voice-Agent

2. Install Python Dependencies

pip install "assemblyai[extras]" ollama elevenlabs

3. Download DeepSeek R1 Model

ollama pull deepseek-r1:7b

4. Configure API Keys

Edit AIVoiceAgent.py and replace the placeholder API keys:

aai.settings.api_key = "YOUR_ASSEMBLYAI_API_KEY"
self.client = ElevenLabs(api_key="YOUR_ELEVENLABS_API_KEY")

🚀 Usage

Start the Voice Agent

python AIVoiceAgent.py

Interaction Flow

1. Speak: The agent listens for your voice input
2. Processing: Your speech is transcribed and sent to DeepSeek R1
3. Response: The AI generates a response (limited to 300 characters for quick interactions)
4. Playback: The response is converted to speech and played back
5. Continue: The conversation continues with maintained context

Stopping the Agent

Press Ctrl+C to stop the voice agent.

⚙️ Configuration

Model Settings

• AI Model: DeepSeek R1 7B (configurable in the code)
• Voice Model: ElevenLabs Turbo v2 (configurable)
• Response Length: Limited to 300 characters (adjustable in system prompt)
• Sample Rate: 16kHz for optimal quality

Customization Options

• Modify the system prompt in AIVoiceAgent.py to change AI behavior
• Adjust response length limits
• Change voice models in ElevenLabs configuration
• Modify audio streaming parameters

🔍 Troubleshooting

Common Issues

"No module named 'assemblyai'"

pip install "assemblyai[extras]"

"Ollama connection error"

• Ensure Ollama is running: ollama serve
• Verify the model is downloaded: ollama list

"Audio device not found"

• Check microphone permissions
• Verify PortAudio installation
• Test microphone with other applications

"ElevenLabs API error"

• Verify API key is correct
• Check API quota/usage limits
• Ensure stable internet connection

Performance Tips

• Use a quality microphone for better transcription accuracy
• Ensure stable internet connection for API calls
• Close unnecessary applications to free up system resources

🏗️ Architecture

┌─────────────────┐    ┌──────────────┐    ┌─────────────────┐
│   Microphone    │───▶│  AssemblyAI  │───▶│   DeepSeek R1   │
│   (Audio Input) │    │ (Speech-to-  │    │ (AI Response    │
└─────────────────┘    │  Text)       │    │  Generation)    │
                       └──────────────┘    └─────────────────┘
                                                      │
┌─────────────────┐    ┌──────────────┐              │
│   Speakers      │◀───│  ElevenLabs  │◀─────────────┘
│ (Audio Output)  │    │ (Text-to-    │
└─────────────────┘    │  Speech)     │
                       └──────────────┘

📄 License

This project is open source. Please check the repository for license details.

🤝 Contributing

Contributions are welcome! Please feel free to submit issues, feature requests, or pull requests.

📞 Support

For issues and questions:

• Open an issue on GitHub
• Check the troubleshooting section above
• Review API documentation for AssemblyAI, Ollama, and ElevenLabs

---

Note: This project requires active internet connection for API services and sufficient system resources to run the DeepSeek R1 model locally via Ollama.