🧠 Knowledge Base Retrieval System (RAG)

A full-stack Retrieval-Augmented Generation (RAG) application that lets users upload private documents (PDFs, DOCX, TXT) and chat with them accurately using AI — powered by Groq LLM and ChromaDB.

---

🚀 Features

• 📄 Upload PDF, DOCX, and TXT documents
• ⚡ Fast AI responses powered by Groq (LLaMA 3.3 70B)
• 🔍 Semantic search using vector embeddings
• 📌 Source citations with every answer (filename, page, excerpt)
• 🌊 Streamed responses word-by-word (like ChatGPT)
• 🗂️ Knowledge base management (upload, view, delete documents)
• 🎯 Confidence scoring based on similarity thresholds
• 🌙 Dark mode support
• 📱 Mobile responsive UI

---

🏗️ Architecture

User Question
     ↓
Embed Query (HuggingFace all-MiniLM-L6-v2)
     ↓
Search ChromaDB → Top 5 Similar Chunks
     ↓
Build Prompt (System + Chunks + Question)
     ↓
Groq API (llama-3.3-70b-versatile)
     ↓
Stream Answer + Return Sources

---

🛠️ Tech Stack

Layer	Technology
Frontend	React + Tailwind CSS
Backend	Node.js + Express
LLM	Groq API (LLaMA 3.3 70B)
Embeddings	HuggingFace all-MiniLM-L6-v2
Vector DB	ChromaDB
PDF Parsing	pdf-parse
DOCX Parsing	mammoth
Streaming	Server-Sent Events (SSE)

---

📁 Folder Structure

knowledge-base-rag/
├── client/                   # React frontend
│   ├── src/
│   │   ├── components/
│   │   │   ├── ChatWindow.jsx
│   │   │   ├── MessageBubble.jsx
│   │   │   ├── SourceCard.jsx
│   │   │   ├── DocumentSidebar.jsx
│   │   │   └── UploadZone.jsx
│   │   ├── pages/
│   │   │   ├── Chat.jsx
│   │   │   └── KnowledgeBase.jsx
│   │   └── App.jsx
│   └── package.json
│
├── server/                   # Express backend
│   ├── routes/
│   │   ├── documents.js      # Upload, list, delete
│   │   └── chat.js           # Query + RAG pipeline
│   ├── services/
│   │   ├── embeddings.js     # HuggingFace embeddings
│   │   ├── vectorDB.js       # ChromaDB operations
│   │   ├── groqLLM.js        # Groq API calls
│   │   └── chunker.js        # Text chunking logic
│   ├── parsers/
│   │   ├── pdfParser.js
│   │   ├── docxParser.js
│   │   └── txtParser.js
│   ├── uploads/              # Temporary file storage
│   └── index.js
│
├── .env.example
├── docker-compose.yml
└── README.md

---

⚙️ Prerequisites

• Node.js v18+
• Python 3.9+ (for ChromaDB)
• npm or yarn
• Groq API Key → console.groq.com

---

🔧 Installation & Setup

1. Clone the repository

git clone https://github.com/your-username/knowledge-base-rag.git
cd knowledge-base-rag

2. Install backend dependencies

cd server
npm install

3. Install frontend dependencies

cd ../client
npm install

4. Install ChromaDB (Python)

pip install chromadb sentence-transformers

5. Configure environment variables

cp .env.example .env

Edit .env and fill in your values:

# Groq API
GROQ_API_KEY=your_groq_api_key_here
GROQ_MODEL=llama-3.3-70b-versatile

# Embeddings
EMBEDDING_MODEL=all-MiniLM-L6-v2

# ChromaDB
CHROMA_HOST=localhost
CHROMA_PORT=8000

# App Config
MAX_FILE_SIZE_MB=20
CHUNK_SIZE=500
CHUNK_OVERLAP=50
TOP_K_RESULTS=5
PORT=3001

6. Start ChromaDB

chroma run --host localhost --port 8000

7. Start the backend server

cd server
npm run dev

8. Start the frontend

cd client
npm run dev

9. Open in browser

http://localhost:5173

---

📡 API Reference

Upload a Document

POST /api/documents/upload
Content-Type: multipart/form-data

Body: file (PDF | DOCX | TXT)

Response:

{
  "id": "doc_abc123",
  "filename": "policy.pdf",
  "chunks": 42,
  "status": "ready"
}

---

List All Documents

GET /api/documents

Response:

[
  {
    "id": "doc_abc123",
    "filename": "policy.pdf",
    "size": "1.2MB",
    "uploadedAt": "2025-06-18T10:00:00Z",
    "chunks": 42,
    "status": "ready"
  }
]

---

Delete a Document

DELETE /api/documents/:id

---

Chat (Query the Knowledge Base)

POST /api/chat
Content-Type: application/json

{
  "question": "What is the refund policy?",
  "documentIds": ["doc_abc123", "doc_xyz456"]
}

Response (streamed via SSE):

{
  "answer": "Based on the uploaded documents, customers can request a refund within 30 days...",
  "sources": [
    {
      "filename": "policy.pdf",
      "page": 2,
      "excerpt": "...eligible for a full refund within 30 days...",
      "score": 0.92,
      "confidence": "high"
    }
  ]
}

---

🎯 Output Confidence Levels

Similarity Score	Confidence	Behavior
0.90 – 1.00	🟢 High	Answer directly
0.75 – 0.89	🟡 Medium	Answer with disclaimer
0.60 – 0.74	🟠 Low	"This might be relevant..."
Below 0.60	🔴 None	"Not found in documents"

---

🐳 Docker Setup (Optional)

docker-compose up --build

Services started:

• Frontend → http://localhost:5173
• Backend → http://localhost:3001
• ChromaDB → http://localhost:8000

---

🧪 Testing

# Backend tests
cd server
npm run test

# Frontend tests
cd client
npm run test

---

🔒 Security Notes

• All uploaded documents are stored locally and never sent to third-party servers (except the text chunks sent to Groq for answering)
• API keys are stored in .env and never exposed to the frontend
• File type and size validation on upload
• Add authentication middleware before deploying to production

---

📌 Supported File Types

Format	Library Used	Max Size
.pdf	pdf-parse	20MB
.docx	mammoth	20MB
.txt	Native Node.js	20MB

---

🗺️ Roadmap

• User authentication (JWT)
• Multi-user support with isolated knowledge bases
• Support for .csv and .xlsx files
• URL/website ingestion
• Document versioning
• Export chat history as PDF
• Ollama support for fully local LLM

---

🤝 Contributing

1. Fork the repository
2. Create a feature branch: git checkout -b feature/my-feature
3. Commit your changes: git commit -m "Add my feature"
4. Push to the branch: git push origin feature/my-feature
5. Open a Pull Request

---

📄 License

MIT License — feel free to use, modify, and distribute.

---

💬 Support

If you run into issues, open a GitHub Issue or reach out at manindrachowdhary1715@email.com.

---

Built with ❤️ using Groq, ChromaDB, and React