AI-driven platform for advanced personalized book recommendations

Overview

The AI Book Manager is an intelligent book management system developed using Python, FastAPI, LangChain and AWS infrastructure. It leverages a PostgreSQL database to store book information and user reviews while incorporating advanced AI features such as generating book summaries and providing personalized recommendations. The core of the AI system is powered by the Llama3 model, deployed on AWS SageMaker, with interaction facilitated via AWS Lambda and API Gateway.

Key Features

1. Database Management: Utilizes AWS RDS PostgreSQL for efficient storage and retrieval of books and reviews, integrated with pgvector for similarity searches.
2. AI-Generated Summaries: Integrates the Llama3 model for generating summaries of books based on their content.
3. Personalized Recommendations: Offers tailored book suggestions using user preferences and reviews.
4. Asynchronous Programming: Implements asynchronous operations for seamless database interactions and AI model predictions.
5. RESTful API: Provides a comprehensive set of endpoints for managing books, reviews and recommendations.
6. Secure Access: Uses JWT authentication to protect API endpoints.
7. Cloud Deployment: Supports deployment on AWS services like EC2, Lambda, or ECS.
8. Dockerization: Includes a Dockerfile for easy containerization and deployment.
9. API Documentation: Offers detailed API documentation through Swagger UI.

RESTful API Endpoints

Book Management

• Add a New Book

  • Method: POST
  • URL: /books
  • Description: Adds a new book to the database.

• Retrieve All Books

  • Method: GET
  • URL: /books
  • Description: Retrieves a list of all books in the database.

• Retrieve a Book by ID

  • Method: GET
  • URL: /books/{id}
  • Description: Retrieves details of a specific book by its ID.

• Update a Book's Information

  • Method: PUT
  • URL: /books/{id}
  • Description: Updates the details of a specific book by its ID.

• Delete a Book by ID

  • Method: DELETE
  • URL: /books/{id}
  • Description: Deletes a specific book by its ID.

Review Management

• Add a Review for a Book

  • Method: POST
  • URL: /books/{id}/reviews
  • Description: Adds a review for a specific book.

• Retrieve All Reviews for a Book

  • Method: GET
  • URL: /books/{id}/reviews
  • Description: Retrieves all reviews for a specific book.

AI Integration

• Get a Summary for a Book

  • Method: GET
  • URL: /books/{id}/summary
  • Description: Retrieves the AI-generated summary for a specific book.

• Generate a Custom Summary

  • Method: POST
  • URL: /generate-summary
  • Description: Generates a summary for provided book content.

Recommendation Engine

• Get Book Recommendations
  • Method: GET
  • URL: /recommendations
  • Description: Provides book recommendations based on user preferences or past reviews.

Detailed Flow Breakdown: Book Creation, Summarization, Recommendation and Login

1. User Login:

   • The user initiates the login process and provides credentials.
   • The system authenticates the user via JWT and issues a token for secure access to the API.
   • The token is required for accessing any protected routes in the application.

2. Book Creation and Summarization:

   • The user sends a request to add a book using the /books endpoint.
   • The system stores the book details in PostgreSQL.
   • The book content is processed by the Llama3 model via AWS API Gateway and Lambda to generate a summary.
   • The generated summary is stored back with embeddings in the database.

3. Book Recommendations:

   • Users request book recommendations based on their preferences or review history.
   • The system retrieves user preferences and performs similarity searches using pgvector embeddings.
   • The system returns a list of recommended books tailored to the user's interests.

This flow ensures efficient handling of book management, summarization and personalized recommendations, all while maintaining secure access and seamless performance through asynchronous operations.

Project Structure

AI-BOOK-MANAGER
 ┣ constants/
 ┃ ┗ app.py
 ┣ database/
 ┃ ┣ __init__.py
 ┃ |
 ┃ ┗ models.py
 ┣ modules/
 ┃ ┣ books/
 ┃ ┃ ┣ routers.py
 ┃ ┃ ┗ schemas.py
 ┃ ┗ users/
 ┃    ┣ routers.py
 ┃    ┗ schemas.py
 ┣ utils/
 ┃ ┣ document_loaders/
 ┃ ┃ ┗ CustomPDFLoader.py
 ┃ ┣ distilbert_embeddings.py
 ┃ ┗ utils.py
 ┣ .env.dev
 ┣ .env.local
 ┣ auth.py
 ┣ Dockerfile
 ┣ log_manager.py
 ┣ main.py
 ┣ Pipfile
 ┣ Pipfile.lock
 ┗ README.md

Implementation Details

Database Schema

• Books Table: id, title, author, genre, year_published, summary, summary_embeddings.
• Reviews Table: id, book_id (foreign key), user_id, review_text, rating.

Llama3 Model Integration

• Deployed on AWS SageMaker.
• Utilizes AWS API Gateway and AWS Lambda for interaction.
• Generates summaries for new book entries and aggregates review summaries.

Asynchronous Programming

• Uses asyncpg and SQLAlchemy with asyncio for asynchronous operations.
• Enhances performance and concurrency in database and AI interactions.

Authentication and Security

• Implements JWT for securing API endpoints.
• Ensures secure database communication.

Dockerization and Deployment

• Dockerfile included for containerizing the application.
• Supports deployment on AWS services using Docker containers.

Additional Features

• AI-Based Recommendations: Utilizes distilbert-base-nli-mean-tokens for embedding generation.
• Text Extraction and Embeddings: Uses the Unstructured library for extracting raw text from book files.

Getting Started

Prerequisites

• Python 3.8+
• Docker
• AWS Account

Installation

1. Clone the repository:

   git clone https://github.com/sk003cs/ai-book-manager.git

2. Navigate to the project directory:

   cd ai-book-manager

3. Build the Docker image:

   docker build -t ai-book-manager .

4. Run the Docker container:

   docker run -p 8000:8000 ai-book-manager

Accessing the API

• API documentation is available at http://localhost:8000/docs via Swagger UI.

Contact

For further assistance or inquiries, please contact SENTHILKUMAR R at [email protected]