🤖 RAG Assistant with Ollama & Qdrant

A Retrieval-Augmented Generation (RAG) assistant that combines local AI models via Ollama with cloud-hosted vector storage using Qdrant Cloud. Built with Streamlit for an intuitive web interface.

🌟 Features

• 📁 Document Processing: Upload and process PDF, TXT, and CSV files with intelligent text extraction
• 🌐 Web Scraping: Extract content from web URLs using BeautifulSoup with smart content detection
• 🔍 Semantic Search: Advanced vector similarity search with session-based filtering
• 💬 Interactive Chat: Natural language querying with context-aware responses and source attribution
• 🏷️ Smart Tagging: Metadata-based document organization and retrieval
• 🔒 Session Isolation: Each session maintains separate document contexts for privacy
• ⚡ Local AI: Fast inference using Ollama with Mistral model (privacy-focused)
• ☁️ Cloud Vector Store: Scalable vector storage with Qdrant Cloud
• 🎨 Modern UI: Clean, responsive interface with loading indicators and visual feedback
• 🚀 Real-time Processing: Live document processing with progress indicators

🏗️ Architecture

┌───────────────────┐    ┌───────────────────┐    ┌─────────────────┐
│    Streamlit UI   │────│  RAG Pipeline     │────│  Qdrant Cloud   │
│  - File Upload    │    │  - Orchestration  │    │  - Vector Store │
│  - Chat Interface │    │  - Session Mgmt   │    │  - Similarity   │
│  - URL Input      │    │  - Error Handling │    │  - Search       │
└───────────────────┘    └───────────────────┘    └─────────────────┘
                                   │
                          ┌──────────────────┐
                          │  Ollama (Local)  │
                          │  - Mistral LLM   │
                          │  - nomic-embed   │
                          │  - Privacy First │
                          └──────────────────┘

🔧 Core Components

1. Document Processor (src/document_processor.py): Handles PDF, TXT, and CSV file processing with multiple fallback methods
2. Web Scraper (src/web_scraper.py): Extracts clean content from web URLs with intelligent content detection
3. Vector Store (src/vector_store.py): Qdrant integration with session-based filtering and metadata indexing
4. LLM Client (src/llm_client.py): Ollama integration for local AI inference with error handling
5. Embedding Client (src/embeddings.py): Generates vector embeddings using nomic-embed-text model
6. RAG Pipeline (src/rag_pipeline.py): Orchestrates the entire retrieval-augmented generation flow

🏷️ Session-Based Architecture

Each document chunk is stored with comprehensive metadata:

{
  "source_type": "document" | "web",
  "source_name": "filename_or_url",
  "session_id": "unique_session_identifier",
  "chunk_id": "chunk_index",
  "content": "actual_text_content"
}

When querying, only vectors matching the current session are retrieved, ensuring complete context isolation between different users or sessions.

🚀 Quick Start

Prerequisites

• Python 3.8+ (Python 3.9+ recommended)
• Ollama installed and running
• Qdrant Cloud account (free tier available - 1GB storage, 100K vectors)
• Git for cloning the repository

📋 Installation Guide

🐧 Linux/macOS Setup

1. Clone the Repository

   git clone https://github.com/yourusername/rag-assistant-ollama.git
   cd rag-assistant-ollama

2. Create and Activate Virtual Environment

   # Create virtual environment
   python3 -m venv venv
   
   # Activate virtual environment
   source venv/bin/activate  # Linux/macOS

3. Install Python Dependencies

   # Upgrade pip first
   pip install --upgrade pip
   
   # Install all required packages
   pip install -r requirements.txt

4. Install and Setup Ollama

   # Install Ollama (if not already installed)
   curl -fsSL https://ollama.ai/install.sh | sh
   
   # Start Ollama service (keep this terminal open)
   ollama serve

5. Download Required AI Models

   # In a new terminal window, download the models
   ollama pull mistral          # Main language model (~4GB)
   ollama pull nomic-embed-text # Embedding model (~274MB)
   
   # Verify models are installed
   ollama list

6. Configure Environment Variables

   # Copy the example environment file
   cp .env.example .env
   
   # Edit the .env file with your Qdrant credentials
   nano .env  # or use your preferred editor

7. Run the Application

   # Make sure your virtual environment is activated
   streamlit run app.py

🪟 Windows Setup

1. Clone the Repository

   git clone https://github.com/yourusername/rag-assistant-ollama.git
   cd rag-assistant-ollama

2. Create and Activate Virtual Environment

   # Create virtual environment
   python -m venv venv
   
   # Activate virtual environment
   venv\Scripts\activate

3. Install Python Dependencies

   # Upgrade pip first
   python -m pip install --upgrade pip
   
   # Install all required packages
   pip install -r requirements.txt

4. Install and Setup Ollama

   • Download Ollama from ollama.ai
   • Install the downloaded executable
   • Open Command Prompt as Administrator and run:

   ollama serve

5. Download Required AI Models

   # In a new Command Prompt window
   ollama pull mistral
   ollama pull nomic-embed-text
   
   # Verify installation
   ollama list

6. Configure Environment Variables

   # Copy the example file
   copy .env.example .env
   
   # Edit .env with Notepad or your preferred editor
   notepad .env

7. Run the Application

   # Ensure virtual environment is activated
   streamlit run app.py

⚙️ Configuration

🔐 Environment Variables Setup

Create a .env file in the project root with the following configuration:

# Qdrant Cloud Configuration (Required)
QDRANT_URL=https://your-cluster-url.qdrant.io
QDRANT_API_KEY=your-qdrant-api-key
COLLECTION_NAME=rag_documents

# Ollama Configuration (Local AI Models)
OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=mistral
EMBEDDING_MODEL=nomic-embed-text

☁️ Qdrant Cloud Setup

1. Create Account: Sign up at Qdrant Cloud
2. Create Cluster:
   • Choose the Free Tier (1GB storage, 100K vectors)
   • Select your preferred region
   • Wait for cluster creation (usually 2-3 minutes)
3. Get Credentials:
   • Copy your Cluster URL (looks like: https://xyz.qdrant.io)
   • Copy your API Key from the cluster dashboard
4. Update Configuration: Add these credentials to your .env file

🤖 Ollama Model Configuration

The application uses two models:

• mistral: Main language model for generating responses (~4GB)
• nomic-embed-text: Embedding model for vector generation (~274MB)

You can change models by updating the .env file:

OLLAMA_MODEL=llama2          # Alternative: llama2, codellama, etc.
EMBEDDING_MODEL=all-minilm   # Alternative embedding models

📖 Usage Guide

🚀 Starting the Application

1. Activate Virtual Environment:

   source venv/bin/activate  # Linux/macOS
   # or
   venv\Scripts\activate     # Windows

2. Start Ollama (in separate terminal):

   ollama serve

3. Launch Application:

   streamlit run app.py

4. Access Interface: Open your browser to http://localhost:8501

📁 Adding Content

Upload Documents

• Supported Formats: PDF, TXT, CSV
• Multiple Files: Upload several files at once
• Auto-Processing: Files are automatically processed when uploaded
• Progress Tracking: Visual progress bar shows processing status

Add Web Content

• Enter URL: Paste any web URL in the input field
• Click "Add URL": Button processes the content
• Smart Extraction: Automatically extracts main content, ignoring navigation and ads
• Loading Indicator: Shows progress while scraping and processing

💬 Chatting with Your Data

1. Ask Questions: Type natural language questions about your content
2. Press Enter: Submit questions using Enter key or the Send button
3. View Sources: See which documents contributed to each answer
4. Session Context: All questions are answered within your current session's context

🔄 Session Management

• Automatic Sessions: Each browser session gets a unique ID
• Clear All: Use the "Clear All" button to start fresh
• Data Isolation: Your documents are never mixed with other users' data

🎯 Example Use Cases

📚 Research Assistant

Upload: Research papers (PDFs), articles (URLs)
Ask: "What are the main findings across these studies?"

📊 Business Intelligence

Upload: Reports (PDFs), company data (CSV)
Ask: "What trends do you see in our quarterly data?"

📖 Learning Companion

Upload: Textbooks (PDFs), online tutorials (URLs)
Ask: "Explain the key concepts from chapter 3"

🔍 Content Analysis

Upload: Multiple documents on a topic
Ask: "Compare the different perspectives presented"

🛠️ Development

📁 Project Structure

rag-assistant-ollama/
├── app.py                    # Main Streamlit application with UI
├── requirements.txt          # Python dependencies
├── .env.example             # Environment variables template
├── .gitignore               # Git ignore patterns
├── README.md                # This comprehensive guide
├── LICENSE                  # MIT License
└── src/                     # Source code modules
    ├── __init__.py          # Package initialization
    ├── rag_pipeline.py      # Main RAG orchestration logic
    ├── vector_store.py      # Qdrant integration and vector operations
    ├── llm_client.py        # Ollama LLM client with error handling
    ├── embeddings.py        # Embedding generation using nomic-embed-text
    ├── document_processor.py # Document processing with multiple formats
    └── web_scraper.py       # Web content extraction and cleaning

🔧 Key Design Decisions

1. Session-Based Isolation: Each user session maintains separate document contexts using unique session IDs
2. Modular Architecture: Clear separation of concerns for maintainability and testing
3. Comprehensive Error Handling: Graceful degradation and user-friendly error messages
4. Scalable Storage: Cloud-based vector storage for production scalability
5. Privacy-Focused: Local AI model inference keeps your data private
6. Responsive UI: Modern, clean interface that works on all devices

🧪 Testing Your Setup

1. Test Ollama Connection:

   curl http://localhost:11434/api/tags

2. Test with Sample Content:

   • Upload a simple text file
   • Add a Wikipedia URL
   • Ask: "What is this content about?"

3. Verify Vector Storage:

   • Check Qdrant Cloud dashboard for stored vectors
   • Verify session isolation by clearing and re-adding content

🚀 Production Deployment

🐳 Docker Deployment

Create a Dockerfile:

FROM python:3.9-slim

WORKDIR /app
COPY requirements.txt .
RUN pip install -r requirements.txt

COPY . .
EXPOSE 8501

CMD ["streamlit", "run", "app.py", "--server.port=8501", "--server.address=0.0.0.0"]

☁️ Cloud Deployment Options

• Streamlit Cloud: Direct deployment from GitHub
• Heroku: Easy deployment with buildpacks
• AWS/GCP/Azure: Full control with container services
• Railway/Render: Simple deployment platforms

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

⭐ If you find this project useful, please consider giving it a star!