🤖 RAG Chatbot for Customer Support

A Retrieval-Augmented Generation (RAG) chatbot that enhances customer support by retrieving relevant documentation and generating accurate responses using Groq's LLaMA-3.1-8B model. The chatbot features an intuitive Streamlit UI and responds strictly based on the provided context.

💡 For queries outside the provided documentation, it replies: "I don't know."

---

🚀 Key Features

• 🔍 Smart Document Retrieval
  Uses Pinecone to fetch the top 3 relevant documents.

• 🤖 Context-Aware Answer Generation
  Powered by Groq's LLaMA-3.1-8B for precise responses.

• ❓ Out-of-Scope Detection
  Answers "I don't know" if information isn't available in the documents.

• 🖥️ Interactive UI
  Built with Streamlit, includes expandable context sections.

• 🧱 Modular Codebase
  Clean separation between UI (interface.py) and backend logic (rag_chain.py).

---

📁 Project Structure

CUSTOMER-SUPPORT-RAG/
├── app/
│   ├── __pycache__/
│   ├── documents.json              # Indexed document data
│   ├── ingest.ipynb                # Notebook for data ingestion
│   ├── interface.py                # Streamlit app
│   └── rag_chain.py                # Core RAG logic
├── 01_data_gathering_logic/
│   ├── angelone_quick_10_links_support_data.json
│   ├── angelone_support_full_data.json
│   ├── insurance_pdfs_flat.json
│   └── Insurance PDFs/
│       └── PDFs/
├── .env                            # Environment variables (GROQ_API_KEY & PINECONE_API_KEY)
├── README.md
├── requirements.txt

---

🛠️ Installation

Install all dependencies:

pip install -r requirements.txt

Included in requirements.txt:

langchain==0.1.10
openai==1.19.0
pinecone
sentence-transformers==2.2.2
PyPDF2==3.0.1
pdfplumber==0.10.2
beautifulsoup4==4.12.3
requests==2.31.0
streamlit==1.33.0
tqdm==4.66.2
numpy==1.26.4
pandas==2.2.2
python-dotenv==1.0.1

---

🔑 Environment Configuration

Create a .env file in the project root:

GROQ_API_KEY=your_groq_api_key_here
PINECONE_API_KEY=your_pinecone_api_key_here

---

🚦 How to Run

1. Ingest Data

Run the ingestion notebook to generate the document index:

open app/ingest.ipynb

• Output: documents.json

2. Launch the Chatbot

streamlit run app/interface.py

3. Start Interacting

• Open: http://localhost:8501
• Ask questions based on the uploaded support documents.
• View supporting documents by expanding the UI panels.

---

🎯 Example Chatbot Behavior

✅ Supported Question

• When the answer is found in the documents:

❌ Unsupported Question

• When the query is not covered:

---

📝 Notes

• Responses are limited to the content in documents.json.
• Ensure documents.json remains in the /app directory.

---

❗ Troubleshooting

• Missing Document Index
  → Re-run ingest.ipynb.

• API Key Not Found
  → Verify .env file exists and contains valid keys.

• Package Errors
  → Confirm all packages are installed with the correct versions.

---

📚 Resources

• LangChain Documentation
• Pinecone Docs
• Streamlit
• Groq API