ritvikneu / talkingagent Goto Github PK

This project implements a simple conversational AI system using Deepgram for speech recognition, Groq for large language model processing, and Langchain for managing conversation history and prompt engineering.

• Real-time Speech Recognition: Utilizes Deepgram's API to transcribe spoken input in real-time.
• Large Language Model Interaction: Integrates with Groq's mixtral-8x7b-32768 model for powerful natural language understanding and generation.
• Conversation Memory: Leverages Langchain's ConversationBufferMemory to maintain conversation context.
• Text-to-Speech: Uses ffplay to convert text responses into spoken output.
• Simple User Interface: Basic command-line interaction, allowing users to speak and receive responses from the AI.

• Python 3.8 or higher: Required for the project's libraries and functionalities.
• Deepgram: Install the Deepgram library: pip install deepgram
• Groq: Install the Groq library: pip install langchain_groq
• Langchain: Install the Langchain library: pip install langchain
• requests: Install the requests library: pip install requests
• ffmpeg: Install the ffmpeg library: sudo apt-get install ffmpeg (Linux) or brew install ffmpeg (macOS)

1. Environment Variables:

   • Create a .env file in the project root directory.
   • Add your API keys:

     DEEPGRAM_API_KEY=YOUR_DEEPGRAM_API_KEY
     GROQ_API_KEY=YOUR_GROQ_API_KEY 
     

   • Obtain API keys from Deepgram and Groq.
   • Important: Replace YOUR_DEEPGRAM_API_KEY and YOUR_GROQ_API_KEY with your actual API keys.

2. System Prompt:

   • The system_prompt.txt file contains the initial instructions for the LLM.
   • You can customize it to influence the AI's behavior and response style.

1. Installation: Install the required dependencies (see "Dependencies" section).
2. Run: Execute the Python script: python conversation.py

• The script will start a live speech recognition session.
• Speak your questions or commands.
• The AI will transcribe your speech, process it with the LLM, and generate a response.
• The response will be spoken back to you via text-to-speech.
• To end the conversation, say "goodbye."

User: Hello, how are you today?
AI: I am doing well, thank you for asking! How can I help you?
User: Can you tell me a joke?
AI: Why don't scientists trust atoms? Because they make up everything! 
User: That's funny! Goodbye.
AI: You have a good one Ritvik! 

• LLM Model: You can change the LLM model by modifying the model_name parameter in LanguageModelProcessor class.
• System Prompt: Adjust the content of the system_prompt.txt file to fine-tune the AI's behavior.
• Text-to-Speech: Experiment with different voices and settings in the TextToSpeech class.

• The project uses ffplay for text-to-speech, so make sure you have ffmpeg installed (see "Dependencies").
• For optimal performance, ensure a stable internet connection and a quiet environment for speech recognition.
• The AI's responses are based on the LLM's capabilities and the provided system prompt.

• Advanced Conversation Management: Implement more sophisticated conversation history and context tracking.
• Multilingual Support: Extend the project to handle multiple languages.
• User Interface: Create a graphical user interface for a more user-friendly experience.
• Domain-Specific Knowledge: Integrate specialized knowledge bases or APIs to enhance the AI's capabilities in specific domains.

This project provides a foundation for building more complex and feature-rich conversational AI systems. You can adapt and extend it to suit your specific needs and creative ideas.

Learn more by doing more, here's where I started @BuildFastWithAI