Banking Support Chatbot

This is an intelligent banking support chatbot that uses sentence embeddings, FAISS indexing, and intent classification to provide accurate responses. It also supports multi-language detection, translation, personalization, and security filtering.

🔧 Features

• Sentence-based semantic search using FAISS
• Intent classification using HuggingFace Transformers
• Multi-language detection + translation
• HTML/CSS/JS based web chat frontend
• Placeholder-based response templating
• Input sanitization and basic data policy enforcement

📁 Project Structure

.
├── app.py                  # Main Flask backend server
├── train.py                # Generate FAISS index & embeddings
├── generate\_csv.py        # Downloads and formats the dataset
├── getDetails.py           # Extracts {{placeholders}} from dataset
├── config.py               # Configuration settings
├── requirements.txt        # Python dependencies
├── .gitignore              # Git ignored files
├── model/
│   ├── embedding\_model.py
│   ├── intent\_classifier.py
│   ├── index\_utils.py
│   └── template\_loader.py
├── utils/
│   ├── text\_utils.py
│   ├── translator.py
│   ├── security\_utils.py
│   └── personalizer.py
├── templates/
│   └── bankingindex.html
├── static/
│   └── chat.js
└── data/
└── (Generated data goes here)

🚀 Getting Started

1. 📦 Install Requirements

Ensure you're using Python 3.8 or newer.

pip install -r requirements.txt

2. 📁 Dataset Preparation

Download and convert the dataset to CSV:

python generate_csv.py

3. 🧠 Train Embeddings + Build FAISS Index

This step encodes the queries, builds FAISS index, and saves embeddings.

python train.py

4. ⚙️ Launch the Flask App

python app.py

Visit: http://127.0.0.1:5000 to interact with the bot.

🧪 Optional Scripts

• Extract placeholders from dataset:

  python getDetails.py
  

• Re-train intent classification model (if needed – currently uses pretrained model):

  Uncomment train_banking77() logic in train.py if required.

✅ To-Do / Improvements

• Add logging and error monitoring (e.g., Sentry)
• Enable user authentication
• Add admin panel for response management
• Deploy using Docker or render.com

🧑‍💻 Author

Developed by Rasikh Ali | Marvelous Software Solutions

📜 License

License MIT