A plugin that adds voice recognition capabilities to Cheshire Cat by transcribing audio files from chat messages and file uploads.

Features

• Transcribe audio from chat messages with audio field
• Process uploaded audio files automatically
• Support for both online (OpenAI API) and offline (Faster Whisper) transcription
• Multiple audio format support: mp3, wav, m4a, mpga, ogg, webm, mpeg, mp4

Requirements

• Cheshire Cat >= 1.8.0
• For local transcription: sufficient disk space for model files
• For online transcription: OpenAI API key

Installation

1. Go to your Cheshire Cat admin panel
2. Navigate to the "Plugin Store" section
3. Search for "Whispering Cat" and install
4. Configure the plugin settings after installation

Configuration

In the settings panel, you can configure:

Basic Settings

• Audio Language: Select the primary language of your audio files
• Use Offline Mode: Toggle between local transcription (Faster Whisper) or online (OpenAI)
• OpenAI API Key: Required only when using online mode

Advanced Settings (Offline Mode)

• Model Size: Select model accuracy vs. speed (tiny, base, small, medium, large)
• Processing Device: Choose between CPU, CUDA (for NVIDIA GPUs), or Auto
• Number of Workers: Set parallel processing threads
• Precision: Balance between accuracy and speed
• Custom Model Path: Optional path for custom models

Usage

Chat Messages with Audio

Send audio in the audio field of your websocket message. Both URL links and Base64 encoded data are supported.

File Uploads

Simply upload audio files in any supported format to have them automatically transcribed and stored in the decalrative memory.

Troubleshooting

• No transcription happening: Check if the plugin is properly configured in settings
• Error messages: Verify your API key if using online mode
• Performance issues: Try a smaller model size or reduce workers if using local mode