📚 FAQ-Bot

FAQ-Bot is a lightweight, semantic search-based chatbot designed to answer frequently asked questions using sentence embeddings. It uses Sentence-Transformers to find the most relevant answer from a set of predefined FAQs.

---

🚀 Features

• ✅ Semantic similarity matching via all-MiniLM-L6-v2 transformer
• ✅ Efficient embedding caching using pickle
• ✅ Simple CLI-based interaction
• ✅ Easily customizable with your own FAQ data
• ✅ Embedding similarity visualization with optional plotting (cosine similarity)

---

🧱 Structure

FAQ-Bot/
├── main.py                 # CLI entry point
├── v1/
│   ├── faq_data.py         # Active FAQ dataset (structured Q&A)
│   ├── faq_bot.py          # Core bot logic and matching
│   └── main.py   
├── v2/
│   ├── faq_data.py         # Embedding creation + caching
│   ├── faq_bot.py          # Updated bot logic and matching
│   ├── viz_utils.py        # Visualization for embeddings (optional)
│   └── main.py        
├── embeddings_cache/
│   └── faq_embeddings.pkl  # Cached embeddings
└── requirements.txt        # Python dependencies

---

📦 Installation

1. Clone the repository:

git clone https://github.com/the-y9/FAQ-Bot.git
cd FAQ-Bot

2. (Optional) Create a virtual environment:

python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

---

▶️ Usage

Start the chatbot via CLI:

python main.py

Type a question into the terminal. The bot will return the best-matching answer using semantic similarity. Type exit or quit to end the session.

---

🧠 How It Works

1. Data loading & embedding:

   • The bot loads FAQ entries from v1/faq_data.py
   • Questions are encoded using a SentenceTransformer model
   • Embeddings are cached for performance (./embeddings_cache/faq_embeddings.pkl)

2. Question Matching:

   • When the user asks a question, it's converted into an embedding
   • The bot computes cosine similarity between the user’s embedding and all cached FAQ embeddings
   • The most similar question is returned, if similarity > 0.3

3. (Optional) Embedding Visualization:

   • If viz_utils.py is configured, the bot can plot the user query and matched question embeddings

---

🛠️ Customize FAQ Data

You can edit or add Q&A entries in v1/faq_data.py. Each entry is structured like:

faq_data = [
    {
        "q": "How do I reset my password?",
        "a": "Click on 'Forgot Password' at login and follow the instructions."
    },
    ...
]

After editing, embeddings will automatically regenerate on next run, as it uses hashing.

---

📊 Requirements

• Python 3.7+
• sentence-transformers
• torch
• numpy
• matplotlib (optional, for visualization)

---

🤝 Contributing

Feel free to fork the repo and submit a pull request!