macos_and_linux_setup.md

Owl - Always-on Wearable AI Setup Guide

<< Home

Prerequisites

Before you begin, ensure you have the following installed on your system:

• Python (version 3.11 or newer)
• Node.js (version 18 or newer)
• Poetry
• FFmpeg
• Ollama (optional, for local-only mode)

Model Agreements

Owl uses PyAnnote for diarization. Please visit Hugging Face and accept the terms for the following models:

• PyAnnote Segmentation Model
• PyAnnote Speaker Diarization Model

Setup Instructions

1. Environment Variables

   Set your Hugging Face token as an environment variable:

   export OWL_ASYNC_WHISPER_HF_TOKEN=<your_hugging_face_token>
   

2. Clone Repository

   Clone the Owl repository from GitHub:

   git clone https://github.com/OwlAIProject/Owl.git
   cd Owl
   

3. Install Dependencies

   Install required OS dependencies

   # on Ubuntu or Debian
   sudo apt install portaudio19-dev python3-pyaudio
   
   # on Mac Os
   brew install portaudio
   

   Use Poetry to install the required dependencies:

   poetry install
   

4. Activate environment Activate virtual env created with poetry

   poetry shell
   

5. Start the Server

   Launch the Owl server:

   owl serve --web
   

   You can now access the web interface at http://localhost:3000. Start testing captures with a microphone or Bluetooth devices. You can also build the iOS app and test captures via the Apple Watch or Bluetooth devices through your iPhone.

Using Commercial Models

If you prefer using commercial models for transcription and summarization, set up the following environment variables instead of the PyAnnote setup:

• For Deepgram:

  export OWL_STREAMING_TRANSCRIPTION_PROVIDER=deepgram
  export OWL_ASYNC_TRANSCRIPTION_PROVIDER=deepgram
  export OWL_DEEPGRAM_API_KEY=<your_api_key>
  

• For GPT-4 Turbo:

  export OWL_LLM_MODEL=gpt-4-turbo-preview
  export OWL_LLM_API_BASE_URL=https://api.openai.com/v1
  export OWL_LLM_API_KEY=<your_api_key>
  

<< Home