Anime Recommender System with LLMOps

An intelligent anime recommendation system that uses OpenAI embeddings and Groq's LLM to provide personalized anime suggestions based on user preferences.

🎯 Features

• Semantic Search: Uses OpenAI embeddings to understand anime descriptions and user queries
• AI-Powered Recommendations: Leverages Groq's Llama 3.1 model for intelligent recommendations
• Vector Database: ChromaDB for efficient similarity search
• Web Interface: Clean Streamlit UI for easy interaction
• Logging & Error Handling: Comprehensive logging and custom exception handling

📋 Prerequisites

• Python 3.10 or 3.11
• OpenAI API key (for embeddings)
• Groq API key (for LLM recommendations)
• UV package manager (recommended) or pip

🚀 Quick Start

1. Clone the Repository

bash git clone <repository-url> cd ANIME-RECOMMENDER-SYSTEM-LLMOPS

2. Install UV (Recommended Package Manager)

```bash

On macOS/Linux

curl -LsSf https://astral.sh/uv/install.sh | sh

On macOS with Homebrew

brew install uv ```

3. Set Up Environment

Create a .env file in the project root with your API keys:

bash OPENAI_API_KEY=sk-your-openai-api-key-here GROQ_API_KEY=gsk_your-groq-api-key-here

4. Install Dependencies

Using UV (recommended): ```bash

Create virtual environment

uv venv --python 3.10

Activate virtual environment

source .venv/bin/activate # On macOS/Linux

or

.venv\Scripts\activate # On Windows

Install dependencies

uv pip install -e . ```

Using pip: ```bash

Create virtual environment

python -m venv .venv

Activate virtual environment

source .venv/bin/activate # On macOS/Linux

or

.venv\Scripts\activate # On Windows

Install dependencies

pip install -e . ```

5. Build the Vector Store

Before running the app, you need to build the vector store with anime embeddings:

```bash

Using default settings

uv run python pipeline/build_pipeline.py

Or with custom options

uv run python pipeline/buildpipeline.py --embed-model-name text-embedding-3-large --persist-dir customdb ```

6. Run the Application

bash uv run streamlit run app/app.py

The app will be available at http://localhost:8501

📁 Project Structure

ANIME-RECOMMENDER-SYSTEM-LLMOPS/ ├── app/ │ ├── __init__.py │ └── app.py # Streamlit web interface ├── config/ │ ├── __init__.py │ ├── config.py # Configuration settings │ └── parse_arguments.py # CLI argument parser ├── data/ │ ├── anime_updated.csv # Processed anime dataset │ └── anime_with_synopsis.csv # Original anime data ├── pipeline/ │ ├── __init__.py │ ├── build_pipeline.py # Build vector store │ └── pipeline.py # Recommendation pipeline ├── src/ │ ├── __init__.py │ ├── data_loader.py # Data loading utilities │ ├── prompt_template.py # LLM prompt templates │ ├── recommender.py # Core recommendation logic │ └── vector_store.py # Vector store management ├── utils/ │ ├── __init__.py │ ├── custom_exception.py # Custom exception handling │ └── logger.py # Logging configuration ├── .env # API keys (create this) ├── pyproject.toml # Project dependencies └── README.md # This file

⚙️ Configuration Options

Embedding Models

You can choose from different OpenAI embedding models:

• text-embedding-3-small (default) - Most cost-effective
• text-embedding-3-large - Higher quality embeddings
• text-embedding-ada-002 - Legacy model

Command Line Options

When building the vector store:

```bash uv run pipeline/build_pipeline.py [OPTIONS]

Options: --embed-model-name TEXT OpenAI embedding model name --persist-dir TEXT Directory to persist vector store --csv-path TEXT Path to CSV file with anime data ```

🔧 Development

Running with Different Embedding Models

```bash

Build vector store with large embeddings

uv run pipeline/build_pipeline.py --embed-model-name text-embedding-3-large

Run the app (it will use the same embedding model automatically)

uv run streamlit run app/app.py ```

Rebuilding the Vector Store

If you update the anime dataset or want to change the embedding model:

```bash

Delete existing vector store

rm -rf chroma_db/

Rebuild with new settings

uv run pipeline/build_pipeline.py ```

🐛 Troubleshooting

Common Issues

1. "OPENAIAPIKEY not found": Make sure your .env file contains valid API keys

2. "command not found: streamlit": Always use uv run prefix when using UV: bash uv run streamlit run app/app.py

3. Deprecation warnings: The warning about Chain.__call__ is expected and doesn't affect functionality

Logs

Check the logs/ directory for detailed application logs if you encounter issues.

📝 License

This project is part of an LLMOps demonstration and is available for educational purposes.

🚢 Deployment

For production deployment using Docker and Kubernetes, see the Deployment Guide.

Security Note: The llmops-k8s.yaml file does not contain secrets. See Kubernetes Secrets Security Guide for handling API keys safely.

🤝 Contributing

Feel free to open issues or submit pull requests for improvements.

📊 Dataset

The anime dataset includes: - Anime titles and descriptions - Genres and ratings - Synopsis information

The system uses this data to provide contextual recommendations based on user preferences.