README.md

metadata

title: App Trial Current
emoji: 📈
colorFrom: purple
colorTo: purple
sdk: gradio
sdk_version: 5.29.1
python_version: 3.12
app_file: app.py
pinned: false
license: mit

Nova Conversation Partner

A real-time, natural-sounding AI conversation partner powered by AWS NovaSonic.

Core Features

• Live, Natural-Sounding AI Conversation: Nova acts as a friendly, responsive conversation partner.
• Low-Latency Real-Time Streaming: Audio streams directly to NovaSonic's bidirectional API.
• Speech Coaching: Analyzes speech patterns and subtly helps improve fluency.
• Session Memory: Continues conversations from where they left off.
• Minimal UI: Simple interface with just a Repeat button and status information.
• Fallback Mode: Works even on systems without audio devices.

Setup

1. Install Dependencies:

   pip install -r requirements.txt
   

2. AWS Credentials: Configure your AWS credentials using environment variables:

   export AWS_ACCESS_KEY_ID=your_access_key
   export AWS_SECRET_ACCESS_KEY=your_secret_key
   export AWS_DEFAULT_REGION=us-east-1
   

   Alternatively, create a .env file in the project root:

   AWS_ACCESS_KEY_ID=your_access_key
   AWS_SECRET_ACCESS_KEY=your_secret_key
   AWS_DEFAULT_REGION=us-east-1
   

Usage

UI Mode (Default)

python app.py

This launches a Gradio web interface with:

• Start/Stop conversation buttons
• Replay last message button
• Status indicator

Command Line Mode

python app.py --cli

In CLI mode, the conversation runs in the terminal with:

• Automatic audio recording/playback
• Press Enter to stop the conversation

Debug Mode

python app.py --debug

Enables detailed logging of API calls and audio processing.

Resume a Session

To continue a previous conversation:

python app.py --session SESSION_ID

Replace SESSION_ID with the ID of a previous session.

Hugging Face Spaces Integration

To use this application in Hugging Face Spaces:

1. Add the following secrets in your Hugging Face Space settings:

   • AWS_ACCESS_KEY_ID
   • AWS_SECRET_ACCESS_KEY
   • AWS_DEFAULT_REGION (optional, defaults to us-east-1)

2. The application will automatically use fallback mode if no audio devices are available.

Configuration

Edit config.py to customize:

• Audio settings
• Greeting message
• Memory limit
• UI text

Troubleshooting

Audio Device Issues

If you encounter the error "Invalid input device (no default output device)":

• The application will automatically switch to fallback mode
• You can still interact with Nova, but no actual audio will be played or recorded
• This is normal in cloud environments and containers without audio hardware

AWS Authentication Issues

If you see authentication errors:

• Ensure your AWS credentials have access to Bedrock services
• Check that you've set the environment variables correctly
• Verify that the region you've specified has access to the Nova Sonic model

Components

The application architecture includes:

• nova_sonic_tool_use.py: AWS Bedrock bidirectional streaming implementation
• language_coach.py: Speech pattern analysis and coaching
• session_manager.py: Conversation persistence and history management
• app.py: Main application logic and UI
• config.py: Configuration settings

License

MIT