EcoSnap

EcoSnap is a Python-based web application for waste classification and sustainability recommendations. It provides both a user-friendly Gradio interface and a FastAPI backend. Users can upload waste images or enter textual queries to classify waste types and receive eco-friendly tips.

Features

• 🖼️ Image Upload & Waste Classification
• 📊 Interactive Gradio UI with Earth-to-Sky theme
• 🌍 Sustainability recommendations via RAG (Retrieval-Augmented Generation)
• 🔌 FastAPI backend for programmatic access
• 🔧 Local FAISS-based vector store with HuggingFace embeddings

Tech Stack

• Python 3.x
• Gradio
• FastAPI
• LangChain & HuggingFace Embeddings
• FAISS vector store
• Pandas, NumPy, Matplotlib
• PIL (Pillow)

Repository Structure

├── app.py            # Gradio UI and front-end logic  
├── model.py          # FastAPI app, RAG setup, and classification API  
├── sustainability_data/
│   └── plastic_general_info.txt  # Sample sustainability docs  
├── faiss_index/      # Saved FAISS index (created at runtime)  
└── requirements.txt  # Python dependencies  

Installation

1. Clone the repository

   git clone https://github.com/thrishank007/EcoSnap.git
   cd EcoSnap

2. Create and activate a virtual environment

   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies

   pip install -r requirements.txt

Usage

1. Run the Gradio UI

python app.py

Open the displayed local URL (e.g., http://localhost:7860) in your browser to interact with the EcoSnap UI.

2. Start the FastAPI Backend

uvicorn model:app --reload

By default, the API will run at http://127.0.0.1:8000. Endpoints:

• POST /classify-image/ – Upload an image to classify waste.
• POST /recommend/ – Send a text prompt to get sustainability tips via RAG.

API docs: http://127.0.0.1:8000/docs

Sustainability Data

On first run, the app generates mock sustainability data under ./sustainability_data/. You can replace these with your own .txt documents:

• Create .txt files in sustainability_data/
• RAG will ingest them and build/update faiss_index/

Contributing

Contributions are welcome! Please open an issue or submit a pull request.

License

This project is licensed under the MIT License. See LICENSE for details.

Contact

Created by @thrishank007.
Feel free to reach out with feedback or questions.