� Simple RAG Assistant

A beginner-friendly Retrieval-Augmented Generation (RAG) application designed for learning and portfolio projects. This project demonstrates the core RAG workflow: document ingestion, text chunking, vector embeddings, semantic search, and AI-powered question answering.

🎯 Learning Objectives

This project is perfect for beginners to understand:

• RAG Fundamentals: How retrieval-augmented generation works
• Vector Embeddings: Converting text to numerical representations
• Semantic Search: Finding similar content using vector similarity
• Document Processing: Extracting and chunking text from various file formats
• Stream Processing: Real-time AI response streaming

---

🏗️ System Architecture

graph TD
    subgraph Streamlit_UI [Streamlit Frontend]
        Upload[Upload Documents]
        Chat[Chat Assistant]
        Search[Semantic Search]
        Summary[Document Summarization]
    end

    subgraph FastAPI_Backend [FastAPI Backend]
        API[REST & Streaming API]
        
        subgraph RAG_Pipeline [RAG Processing Pipeline]
            DocProc[Document Processor]
            Chunker[Text Chunker]
            Embedder[Sentence Transformers]
            VectorDB[ChromaDB Vector Store]
            LLM[Gemini 1.5 Flash]
        end
    end

    subgraph Local_Storage [Local Storage]
        Files[Uploaded Documents]
        Chroma[ChromaDB Database]
    end

    Upload --> API
    Chat --> API
    Search --> API
    Summary --> API
    API --> DocProc
    API --> VectorDB
    API --> LLM
    DocProc --> Chunker
    Chunker --> Embedder
    Embedder --> VectorDB
    VectorDB --> LLM
    Files --> DocProc
    VectorDB --> Chroma

Loading

---

🚀 Features

Core RAG Workflow

• Document Upload: Support for PDF, DOCX, PPTX, and TXT files
• Text Extraction: Page-by-page text extraction from documents
• Text Chunking: Intelligent text splitting with overlap for context preservation
• Vector Embeddings: Converting text chunks to vector embeddings using sentence-transformers
• Semantic Search: Finding similar content using cosine similarity in ChromaDB
• AI Q&A: Answering questions using retrieved context with Gemini API
• Document Summarization: Generating structured summaries of uploaded documents

Technical Stack

• Frontend: Streamlit (Python web framework)
• Backend: FastAPI (Python web framework)
• Vector Database: ChromaDB (local vector store)
• Embeddings: sentence-transformers (all-MiniLM-L6-v2)
• LLM: Google Gemini 1.5 Flash API
• Document Processing: pypdf, python-docx, python-pptx

---

📋 Prerequisites

• Python 3.8 or higher
• Google Gemini API Key (free tier available)
• pip (Python package manager)

---

🔧 Installation

1. Clone the Repository

git clone <repository-url>
cd DocQuery

2. Set Up Backend

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

pip install -r requirements.txt

3. Configure Environment Variables

Create a .env file in the backend directory:

GEMINI_API_KEY=your_gemini_api_key_here

Get your free API key from: https://makersuite.google.com/app/apikey

4. Set Up Frontend

cd frontend
python -m venv env
env\Scripts\activate  # On Windows
source env/bin/activate  # On Mac/Linux

pip install -r requirements.txt

5. Configure Frontend Environment

Create a .env file in the frontend directory:

BACKEND_URL=http://127.0.0.1:8000

---

🎮 Usage

Start the Backend Server

cd backend
venv\Scripts\activate  # On Windows
python -m backend.app.main

The backend will start at: http://127.0.0.1:8000

Start the Frontend

cd frontend
env\Scripts\activate  # On Windows
streamlit run app.py

The frontend will open in your browser at: http://localhost:8501

---

📖 How to Use

1. Upload Documents

• Navigate to Upload Documents in the sidebar
• Select PDF, DOCX, PPTX, or TXT files
• Click Process & Upload to ingest documents
• The system will extract text, chunk it, and create vector embeddings

2. Chat with Documents

• Navigate to Chat Assistant
• Ask questions about your uploaded documents
• The AI will retrieve relevant context and provide answers with source citations

3. Semantic Search

• Navigate to Semantic Search
• Enter a search term or concept
• View semantically similar passages with relevance scores

4. Document Summarization

• Navigate to Summaries
• Select a document from your library
• Generate an AI-powered summary with key insights

---

🧠 How RAG Works

This project implements the standard RAG pipeline:

1. Document Ingestion: Upload documents (PDF, DOCX, PPTX, TXT)
2. Text Extraction: Extract text page-by-page from documents
3. Text Chunking: Split text into smaller chunks (1000 chars with 200 overlap)
4. Embedding Generation: Convert each chunk to a vector using sentence-transformers
5. Vector Storage: Store embeddings in ChromaDB with metadata
6. Query Processing: Convert user query to vector embedding
7. Semantic Search: Find similar chunks using cosine similarity
8. Context Retrieval: Retrieve top-K most relevant chunks
9. Prompt Construction: Build prompt with retrieved context
10. AI Generation: Generate answer using Gemini API with retrieved context

---

📁 Project Structure

DocQuery/
├── backend/
│   ├── app/
│   │   ├── main.py              # FastAPI application entry point
│   │   ├── api/
│   │   │   └── endpoints.py    # API endpoints
│   │   └── services/
│   │       ├── doc_processor.py    # Document text extraction
│   │       ├── vector_store.py     # ChromaDB vector operations
│   │       └── rag_service.py      # RAG logic & LLM integration
│   ├── requirements.txt        # Backend dependencies
│   └── .env                   # Backend environment variables
├── frontend/
│   ├── app.py                 # Streamlit application entry point
│   ├── pages/
│   │   ├── upload_page.py     # Document upload interface
│   │   ├── chat_page.py       # Chat assistant interface
│   │   ├── search_page.py     # Semantic search interface
│   │   └── summary_page.py    # Document summarization interface
│   ├── services/
│   │   └── api_client.py      # Backend API client
│   ├── components/
│   │   └── styles.py          # UI styling
│   └── requirements.txt        # Frontend dependencies
└── sample_docs/
    └── sample_policy.txt      # Sample document for testing

---

🔑 Key Concepts for Interviews

Vector Embeddings

• Converting text to numerical vectors that capture semantic meaning
• Similar concepts have similar vectors in high-dimensional space
• Used for semantic search instead of keyword matching

ChromaDB

• Open-source vector database for storing and querying embeddings
• Supports cosine similarity search
• Stores vectors with metadata for filtering

RAG (Retrieval-Augmented Generation)

• Combines retrieval systems with LLMs
• Retrieves relevant context before generating answers
• Reduces hallucinations by grounding responses in retrieved data

Sentence Transformers

• Pre-trained models for generating text embeddings
• all-MiniLM-L6-v2 is fast and efficient for general use
• Maps text to 384-dimensional vectors

Text Chunking

• Splitting long documents into smaller, manageable pieces
• Preserves context with overlapping chunks
• Improves retrieval accuracy for specific information

---

🛠️ Technologies Used

• Streamlit: Frontend web framework
• FastAPI: Backend web framework
• ChromaDB: Vector database
• sentence-transformers: Text embedding model
• Google Gemini: Large Language Model API
• pypdf: PDF text extraction
• python-docx: DOCX text extraction
• python-pptx: PPTX text extraction

---

📝 Notes for Beginners

• This project uses in-memory document storage for simplicity
• No authentication or user management (simplified for learning)
• No conversation history persistence (session-based only)
• Uses semantic search only (no keyword search)
• Designed for local development and learning purposes

---

🚀 Future Enhancements (Learning Path)

After mastering this project, consider adding:

• User authentication with JWT
• Conversation history with database storage
• Hybrid search (semantic + keyword)
• Document comparison features
• Report generation
• Docker containerization
• Deployment to cloud platforms

---

📄 License

This project is open source and available for educational purposes.

---

🤝 Contributing

This is a learning project. Feel free to fork, modify, and use it for your portfolio!