Skip to content

Installation

Requirements

  • Python 3.10 or higher
  • Docker is required to run this project. You can download it here.
  • ffmpeg is required to run this project. You can download it here.
  • Redis is required to run this project. You can download it with sudo apt install redis.
  • Hugging Face token is required to use the Hugging Face models. You can get a token here. You must also accept the terms and conditions for the models you want to use (here and here).

Installation

  1. Clone the repo
git clone https://github.com/TCU-ClassifAI/classifAI-engine.git
cd classifAI-engine
  1. Install Python packages (it is recommended to use a virtual environment)
pip install -r src/requirements.txt
  1. Install required system packages 

# Install ffmpeg
sudo apt install ffmpeg # Ubuntu
brew install ffmpeg # MacOS
# install Redis
sudo apt install redis-server # Ubuntu
brew install redis # MacOS
# start Redis
sudo redis-server # Ubuntu
4. Accept the Hugging Face terms and conditions - 1. visit hf.co/pyannote/speaker-diarization-3.1 and accept user conditions - 2. visit hf.co/pyannote/segmentation-3.0 and accept user conditions - 3. visit hf.co/settings/tokens to create an access token

  1. Set up the environment variables

  2. For a full list of environment variables, see the Configuration documentation. This guide will cover a minimal set of environment variables to get the server running.

  3. Create a .env file in the root directory of the project. The .env file should contain the following environment variables:

# .env
HF_TOKEN=your_hugging_face_token # Get a token from https://hf.co/settings/tokens
REDIS_PORT=6379 #(default port for Redis, can be changed), the URL can also be changed (see config)
LLAMA_API_URL=http://localhost:5003 # URL for the LLAMA API
5. (Optional) Set up configuration for the API

  • Edit values in src/config/config.py to match your environment.

  • Launch the API

python src/app.py
  1. Run your RQ worker (you can do this through supervisor or another process manager)
rq worker -c config.worker_config

Test

curl http://localhost:5000/healthcheck should return OK

More usage can be found in the API Documentation.

Installation on the GPU Server

Currently, it is run as a Systemd service. The service file is located at /etc/systemd/system/classifai.service. The service can be started, stopped, and restarted using the following commands:

sudo systemctl start classifai
sudo systemctl stop classifai
sudo systemctl restart classifai

The logs can be viewed using the following command:

sudo journalctl -u classifai