AI Chatbot with FAISS & Hugging Face

Conversational AI Chatbot powered by FAISS for retrieval-based responses and Hugging Face embeddings. This chatbot provides human-like responses using a customer service policy knowledge base.

---

Features

Retrieval-Augmented Generation (RAG)** with FAISS.
Hugging Face embeddings for similarity search. Conversational responses** (not just raw text dumps).
Handles queries on orders, refunds, shipping, and more.**
Multi-turn conversation support.

---

Installation

1️⃣ Clone the Repository

git clone https://github.com/yourusername/AI-Chatbot.git
cd AI-Chatbot

2️⃣ Create a Virtual Environment

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

3️⃣ Install Dependencies

pip install -r requirements.txt

4️⃣ Set Up API Keys

Create a .env file and add:

HUGGINGFACE_TOKEN=your_huggingface_api_key

Or set it as an environment variable:

export HUGGINGFACE_TOKEN=your_huggingface_api_key  # On Windows: set HUGGINGFACE_TOKEN=your_huggingface_api_key

---

Project Structure

📦 AI-Chatbot
 ┣ 📂 Company_policy.txt        # Stores the company policy text
 ┣ 📂 faiss_index              # FAISS storage
 ┣ 📜 config.py                # Configuration settings
 ┣ 📜 policy_loader.py         # Loads & processes company policies into FAISS
 ┣ 📜 rag_pipeline.py          # Main RAG model handling FAISS queries
 ┣ 📜 main.py                  # Runs the chatbot in interactive mode
 ┣ 📜 requirements.txt         # Dependencies
 ┣ 📜 README.md                # Project Documentation

---

How to Use

1️⃣ Index the Policy Data

Before using the chatbot, process & store the company policies into FAISS:

python policy_loader.py

2️⃣ Run the Chatbot

python main.py

You'll see:

✅ FAISS Index loaded successfully!
✅ Chatbot is ready! Type 'exit' to quit.
You: When will my order arrive?
Chatbot: Your order should arrive in 5-7 business days for standard shipping. Would you like me to check your tracking details?

---

Troubleshooting

Problem	Solution
FAISS index not found	Run policy_loader.py to generate the FAISS index.
Hugging Face API key missing	Set HUGGINGFACE_TOKEN in .env or as an environment variable.
Chatbot gives generic responses	Check policy_loader.py to confirm policies were indexed correctly.

---

Next Steps

Improve Conversational Flow by integrating GPT-4 for better responses.
Add a Web UI using FastAPI & React for real-time interactions.
Log User Queries to improve responses over time.