🧠 ComicFinder

ComicFinder is an AI-powered content-based recommendation system built using Python and OpenAI Embeddings. It helps users discover semantically similar manga, manhwa, manhua, and webtoons based on natural language descriptions, genres, or titles — ideal for fans seeking personalized recommendations beyond keyword search.

---

💻 Live Demo Of Comic Finder

https://comicfinder.streamlit.app/

---

🚀 Features Of Comic Finder

• 🔍 Recommends similar manga/manhwa/manhua/webtoon based on descriptions or titles
• 📦 Utilizes precomputed clean_embeddings.npy for fast results
• 🧠 Embedding generation using OPENAI embeddings api
• ⚡ Fast cosine similarity search for real-time recommendation
• 🖥️ Clean Streamlit-based frontend
• 📁 Organized data and scripts for easy retraining or extension

---

📁 Project Structure

comic-recommender/
├── app.py                       # Main application script
├── data/
│   ├── data.csv                 # Original manhwa dataset
│   ├── clean_data.csv           # Cleaned and preprocessed data
│   └── clean_embeddings.npy     # (Ignored from Git, must be downloaded separately)
├── scripts/
│   ├── clean_dataset.py         # Data cleaning script
│   ├── generate_embeddings.py   # Embedding generation
│   └── recommend.py             # Similarity-based recommendations but CLI version
├── .env                         # Store API keys 
├── requirements.txt             # Python dependencies
└── README.md                    # You're here!

---

🔧 How to Install and Run ComicFinder Locally

git clone https://github.com/AdityaEXP/ComicFinder.git
cd ComicFinder

# Optional: Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

pip install -r requirements.txt
streamlit run .\app.py

---

📌 Example Use Cases

• Find romance manhwa similar to What's Wrong with Secretary Kim?
• Get fantasy webtoon recommendations with strong male leads
• Discover hidden manga gems with character development arcs
• Replace genre filters with AI-powered natural language queries

---

📥 Download Embedding File

Since clean_embeddings.npy is large, it’s not included in this repo. 📦 Download clean_embeddings.npy Or you can also generate the clean_embeddings.npy using your own openai api key it will cost around $0.02 per generation

---

🔐 Environment Variables

Create a .env file for your OpenAI API Key

OPENAIKEY=sk-xxxxxx

---

📜 License

MIT — free to use, modify, and distribute.

---

🤝 Author

Aditya 🛠️ AI + Python + Web3 Enthusiast

---

🔮 Future Plans

1. Replace cosine similarity by FAISS for fast searches
2. Adding Anime and webseries dataset as well
3. Create a automated source for scrapping data from webpages or api and update the dataset periodically.
4. Improve searching by creating high value embeddings using more data etc.

---

📚 Dataset Source and Preprocessing

This project uses data inspired by or adapted from the following Kaggle dataset:

📊 Kaggle - Manhwa and Webtoon Dataset
Credit to Victor Soeiro for compiling and sharing this dataset.