Custom RAG System with Ollama

A flexible Retrieval-Augmented Generation (RAG) system built with Ollama, designed for efficient document processing and contextual query responses.

Features

• 📄 Multi-format document support (PDF, DOCX, Markdown, TXT)
• 🔍 Advanced vector similarity search
• 🤖 Integration with Ollama's language models
• 📊 Built-in benchmarking capabilities
• 🎯 High-precision context retrieval

Getting Started

Prerequisites

• Python 3.8+
• Ollama installed and running
• Required Python packages (specified in requirements.txt)

Installation

# Clone the repository
git clone https://github.com/yourusername/custom-rag-ollama.git

# Navigate to project directory
cd custom-rag-ollama

# Install dependencies
pip install -r requirements.txt

Usage

1. Document Upload

• Access the Document Upload section in the UI
• Supported file formats:
  • PDF
  • DOCX
  • Markdown
  • Plain text
• Documents are automatically processed and chunked for optimal retrieval

2. Query Interface

• Navigate to the Query tab
• Input your question
• Configure parameters:
  • Number of results to retrieve
  • Ollama model selection
• View results with:
  • Generated response
  • Source document metadata
  • Similarity scores

3. Benchmarking

Run performance tests to evaluate:

• Embedding generation speed
• Search efficiency
• Model response quality
• Overall system performance

System Architecture

Core Components

1. Document Processing Pipeline

   • Text extraction from multiple formats
   • Intelligent document chunking
   • Metadata preservation

2. Embedding Engine

   • Vector embedding generation
   • Optimization for search performance
   • Model-agnostic design

3. Vector Search System

   • High-performance similarity matching
   • Configurable search parameters
   • Result ranking optimization

4. Response Generation

   • Integration with Ollama LLMs
   • Context-aware response synthesis
   • Source attribution

Contributing

We welcome contributions! Please feel free to submit a Pull Request.

License

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

Acknowledgments

• Ollama team for their excellent LLM framework
• Contributors and community members

Contact

For questions and support, please open an issue in the GitHub repository.

---