playlist2vec

ABOUT

This repository contains code for playlist2vec.com, a website built to demo the vector search model described in our paper, "Representation, Exploration, & Recommendation of Playlists".

FEATURES

• Intuitive search interface and Spotify integration
• Microservices architecture for seamless scalability.
• Docker containers for a streamlined deployment.
• Utilization of low-cost libraries (SQLite and Usearch Vector Search) to minimize app footprint.
• MMAP-based vector search enabling the application to operate efficiently on budget-friendly machines such as Raspberry Pi.
• NGINX configuration for a robust traffic handling.
• Rate limiting implemented to safeguard against DDoS attacks.
• NGINX caching to optimize server resource usage.
• Auto-scaling capabilities using Kubernetes.

INSTALLATION

Note: This setup has been tested on Ubuntu 22.04 for both x86_64 and aarch64 architectures.

1. Install Docker

1. Follow the instructions to install Docker from this link: Install Docker On Ubuntu.
2. Make sure to complete the post-installation steps outlined here.

2. Install Nginx

Install Nginx by following the guide available at: How to Install Nginx on Ubuntu 22.04.

3. Install NodeJS

You can install Node.js (v20.18.0) by referring to this tutorial: How to install Node.js on Ubuntu 22.04.

4. Download Code

Run the following command to clone the repository:

git clone https://github.com/piyp791/playlist2vec.git

5. Navigate To Project

Navigate to the project directory by running:

cd playlist2vec

6. Nginx Setup

1. Copy the nginx/nginx.conf file to /etc/nginx/.

2. Copy the nginx/site_config file to /etc/nginx/sites-available/<YOURSITENAME>.

3. Create a symbolic link from sites-available to sites-enabled with the following command:

   sudo ln -s /etc/nginx/sites-available/<YOURSITENAME> /etc/nginx/sites-enabled/
   

4. Remove the default configuration file by executing:

   sudo rm /etc/nginx/sites-enabled/default
   

5. Create a cache directory for NGINX:

   sudo mkdir /var/cache/nginx
   

6. Verify the configuration is correct by running:

   sudo nginx -t
   

7. Restart NGINX with the command:

   sudo systemctl restart nginx
   

7. Install the Application

From within the project directory, execute the build script:

./build.sh.

This will create an HTTP version of the website, which can be integrated with a service like a Cloudflare tunnel for an HTTPS frontend.

Note

By default, the build script configures the application to use the mini version of the corpus, which contains 377,000 items. In this configuration, the search index for the mini corpus is loaded into memory. In contrast, the search index built with full corpus is accessed in memory-mapped mode.

To change this setting, you can modify the .env file. Set the IS_MINI variable to false to use the full version of the corpus:

IS_MINI=false

RUN IN DEVELOPMENT MODE (WITHOUT NGINX)

Setting up Search Service:

cd search-service
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
uvicorn src.server:app --port 3001

Setting up Autocomplete Service:

cd autocomplete-service
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
uvicorn src.server:app --port 3002

Setting up Web-Server:

cd web-server
npm install
npm run start-dev-mode

RUN TESTS

cd web-server
npm run test-dev

CITATION

Papreja, P., Venkateswara, H., Panchanathan, S. (2020). Representation, Exploration and Recommendation of Playlists. In: Cellier, P., Driessens, K. (eds) Machine Learning and Knowledge Discovery in Databases. ECML PKDD 2019. Communications in Computer and Information Science, vol 1168. Springer, Cham. https://doi.org/10.1007/978-3-030-43887-6_50