mhuseynli / infreco

Event-driven infrastructure for Recommender Systems built with Python(Django)

Documentation for Event-Driven Recommender System

Overview

This document provides a comprehensive overview of the Event-Driven Recommender System developed for webshops. The system's core functionality includes the registration of webshops, capturing user interactions as events, processing these events via RabbitMQ for real-time data handling, and providing a recommender engine interface for personalized user recommendations.

Key Components

1. Webshop Registration and Management: Webshops can register, update their information, and be deleted from the system.
2. Event Handling: Capturing user interaction events (clicks, views, etc.) from webshops.
3. RabbitMQ Integration: Utilizing RabbitMQ for queuing and processing events asynchronously.
4. Recommender Engine Interface: An abstract interface for integrating various recommender algorithms.
5. Data Models: MongoDB data models for webshops, users, items, and events.

Installation and Setup

1. MongoDB: Ensure MongoDB is installed and running.
2. RabbitMQ: Install and configure RabbitMQ. Ensure the service is running.
3. Python Environment: The system is developed in Python. Ensure Python and necessary libraries (pika, Django, djangorestframework) are installed.

Usage

Webshop Registration

Webshops can register using the /webshops/register/ endpoint. Required data includes email, webshop name, and password.

Event Capturing

Webshops send user interaction events to /events/ with the X-API-KEY header for authentication. The event data should include user and item details along with the event type and timestamp.

Event Processing with RabbitMQ

Events sent to the /events/ endpoint are queued in RabbitMQ and processed asynchronously. The processing logic includes user and item data updates in MongoDB.

Recommender Engine Interface

The abstract interface RecommenderInterface in recommender_interface.py allows integration of various recommendation algorithms. It includes methods for training, generating recommendations, and updating models.

API Endpoints

• POST /webshops/register/: Register a new webshop.
• GET /webshops/: Retrieve webshop details (requires X-API-KEY).
• PUT /webshops/: Update webshop details (requires X-API-KEY).
• DELETE /webshops/: Delete a webshop and related data (requires X-API-KEY).
• POST /events/: Send user interaction events for processing.
• POST /train/: Initiate training of the recommender model (requires X-API-KEY).
• GET /recommend/{user_id}/: Get recommendations for a user (requires X-API-KEY).

Data Models

Webshop

• Email
• Name
• Password (hashed)

User

• Webshop ID
• External ID
• Attributes

Item

• Webshop ID
• External ID
• Attributes

Event

• Webshop ID
• User ID
• Item ID
• Event Type
• Timestamp

RabbitMQ Consumer

The consume.py script listens for new messages in the webshop_events queue and processes them by updating user and item information and storing the events in MongoDB.

Recommendations

The RecommendationView handles requests for user recommendations. It utilizes the recommender algorithm defined in MyRecommender, which adheres to the RecommenderInterface.

Security

• Webshop API key authentication.
• Passwords are stored in a hashed format.

Future Enhancements

• Implementation of various recommender algorithms.
• Dashboard for webshops to manage their data and view analytics.
• Enhanced error handling and logging mechanisms.

---

This documentation serves as a guide for developers and users interacting with the Event-Driven Recommender System. Further details can be found in the codebase and additional documentation.