amen-server

Server to run Amen analysis and return JSON results.

Running Locally

• Install amen
• Add keys and secrets to uploaders/s3.py
• Create your buckets and set them to public (link to aws doc)
• Build the dockerfile: docker build -t amen-server-test .
• Run the dockerfile: docker run -p 4000:80 amen-server-test
• Test with CURL: curl -X POST -F "[email protected];type=audio/mpeg" http://localhost:4000
• Stop docker: docker stop $(docker ps -q)

Running on AWS

• Create a new machine on AWS! (We recommend Ubuntu Server 16.04)

• Make sure port 80 is open to everyone.

• Get the server code on your machine - we recommend checking it out from Github.

• Create your buckets, set them to public, and update them in uploaders/s3.py

• Add your AWS keys and secrets to uploaders/s3.py

• Update the path for the Tornado webserver in server.py: we recommend /amen-server, as that's what the below nginx config uses.

• Set up NGINX with proper ports and routes, and get it running:

  • sudo apt-get update
  • sudo apt-get install nginx
  • Add the following lines to /etc/nginx/sites-enabled/default, under listen [::]:80 default_server;:

    location /amen-server {
        proxy_pass http://localhost:4000/amen-server;
        client_max_body_size 8M;
    }
  

  • start nginx sudo service nginx start

• Install Docker: sudo apt-get install docker.io

• Add the current user (ubuntu) to the docker group, so we can run without sudo: sudo gpasswd -a $USER docker

• Build the dockerfile: docker build -t amen-server-test .

• Run the dockerfile: docker run -p 4000:80 amen-server-test

• Send a test CURL: curl -X POST -F "[email protected];type=audio/mpeg" http://<your-aws-url>/amen-server

Some Notes On Performance

Amen is computationally heavy, and is a Python library. It could be faster. The mean processing time for a 3:30 audio track on a AWS free-tier t2.micro is 20.5 seconds - that drops to 16.4 seconds on an AWS c4 machine. Plan accordingly.