CariConnect: Ranked Recommendation System

CariConnect is a recommendation system that matches upcoming authors with prospective publishers using a combination of clustering, cosine similarity, and Retrieval-Augmented Generation (RAG) techniques. This project integrates advanced machine learning models and LLMs for personalized recommendations and detailed explanations.

Features

• Spectral Clustering: Groups books into clusters using BERT embeddings.
• Cosine Similarity: Ranks similar books within clusters.
• Retrieval-Augmented Generation (RAG): Enhances matches and provides explanations using LangChain and LLAMA LLM.
• An interactive web interface for user input and results.

Prerequisites

• Obtain API keys for HuggingFace and Groq.
• Add keys to a .env file in the project root.

Setup Instructions

1. Clone the Repository

git clone https://github.com/khantnhl/cari-connect.git
cd cari-connect

2. Install Dependencies

pip install -r requirements.txt

3. Run the Application:

run app.py
#Access at: http://127.0.0.1:8000

4. Optional: To retrain the Spectral Clustering model, use the provided notebook:

jupyter notebook Generate_Spectral_Clustering_Model_Files.ipynb

Team: Ashley Camacho-Medellin, Khant Nyi Hlaing, Kaylin (Kienn) Nguyen, Eileen (Yiming) Xue, Grace (Yunjin) Zhu TAs: Arjun Aggarwal, Rebecca Aurelia Dsouza Advisors: V. Steve Russell, Solomon Perkins

License This project is licensed under the MIT License.