✈️ A320 AI Assistant RAG Chatbot

A Retrieval-Augmented Generation (RAG) based AI chatbot built using Django, Pinecone, HuggingFace Embeddings, and Groq LLM.

This system allows users to query a technical PDF document (A320 Aircraft Characteristics – Airport & Maintenance Planning) and receive context-grounded answers.

🚀 Features

• 📄 PDF-based knowledge system
• 🔍 Semantic search using vector embeddings
• 🧠 Context-aware AI responses (RAG architecture)
• ⚡ Fast LLM inference using Groq (Llama 3.1)
• 🎨 Modern animated UI (Glassmorphism design)
• 🔐 Secure API key management via .env

---

🏗️ Architecture

Frontend (HTML/JS)                         User Question
        ↓                                        ↓
Django API (/ask)                          Django Backend
        ↓                                        ↓
Embed Query (BGE)                    SentenceTransformer Embedding
        ↓                                        ↓
Pinecone (Top-k chunks)                  Pinecone Vector Search
        ↓                                        ↓
Construct Prompt                          Retrieve Top Chunk
        ↓                                        ↓
Gemini API                               Groq LLM (Llama 3.1)
        ↓                                        ↓
Return Answer                           Context-Grounded Answer

---

🧠 How RAG Works in This Project

1. PDF is parsed and chunked (in Google Colab).
2. Each chunk is converted into embeddings.
3. Embeddings are stored in Pinecone.
4. When user asks a question:
   • Question is embedded.
   • Most relevant chunk is retrieved.
   • Retrieved context is sent to Groq LLM.
   • LLM generates answer strictly from context.

🛠️ Tech Stack

Layer	Technology Used
Backend	Django
Embeddings	BAAI/bge-small-en-v1.5
Vector DB	Pinecone (Free Tier)
LLM	Groq (Llama 3.1-8b-instant)
Frontend	HTML + CSS + JS
Deployment	Render (Planned)

---

📦 Installation

1️⃣ Clone Repository

git clone https://github.com/Yash-Lade/RAG_chatbot
cd rag_backend

2️⃣ Create Virtual Environment

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

3️⃣ Install Dependencies

pip install -r requirements.txt

If requirements.txt not created, install manually:

pip install django
pip install sentence-transformers
pip install pinecone-client
pip install groq
pip install python-dotenv

🔐 Environment Variables

Create a .env file in project root:

PINECONE_API_KEY=your_pinecone_key
GROQ_API_KEY=your_groq_key

Add .env to .gitignore:

.env
venv/
__pycache__/

▶️ Running the Project

python manage.py migrate
python manage.py runserver

Open browser:

http://127.0.0.1:8000/

📂 Project Structure

rag_backend/
│
├── manage.py
├── .env
│
├── rag_backend/
│   ├── settings.py
│   ├── urls.py
│
└── chatbot/
    ├── views.py
    ├── rag_utils.py
    ├── urls.py
    ├── templates/
    │     └── index.html
    └── static/
          └── style.css

🧪 Sample Test Questions

• What is the maximum ramp weight of the A320-200?
• Describe the jacking procedure for maintenance.
• What wind limits apply during aircraft jacking?
• What protections are provided by the engine nacelle?

Out-of-scope questions should return:

"Not available in document."

⚙️ Configuration Details

Embedding Model

BAAI/bge-small-en-v1.5
Dimension: 384
Similarity: Cosine

LLM Model

llama-3.1-8b-instant

Retrieval Strategy Top K = 1

Context limited to 1200 characters

Max output tokens = 300