PaperBrain

PaperBrain

PaperBrain is an intelligent research paper Q&A system that combines vector search and large language models to provide context-aware answers to research-related questions. It processes academic papers, understands their content, and generates structured, informative responses with proper citations and context.

AI Response

Key Features

Core Capabilities

• Smart Vector Search: Utilizes Qdrant for semantic similarity search of research papers
• Intelligent Analysis: Leverages LLaMA 3.2 for generating comprehensive, context-aware answers
• Structured Responses: Provides organized output with:
  • Main answer summary
  • Key points from papers
  • Paper citations and references
  • Analysis limitations
• Duplicate Detection: Intelligent tracking of shown papers to avoid repetition

Paper Citations

Paper Keybopints

Advanced Features

• Analytics Dashboard: Track system usage, search patterns, and relevance metrics
• Conversation History: Maintain records of previous queries and responses
• Relevance Scoring: Clear explanation of paper matching with detailed relevance metrics
• Interactive Commands: System controls for analytics, history, and paper tracking

Technology Stack

• Vector Store: Qdrant for efficient similarity search
• Embeddings: Nomic Embed Text for paper vectorization
• LLM Integration: LLaMA 3.2 (1B parameter model) via Ollama
• Infrastructure: Docker containerization
• Backend: Async Python with modern libraries
• API Layer: Async HTTP with HTTPX

Qdrant Database

Markdown to Vectors

? Prerequisites

# System requirements
- Python 3.9+
- Docker
- 4GB+ RAM for LLM operations
- Disk space for paper storage

Installation

1. Clone the repository:

git clone https://github.com/ansh-info/PaperBrain.git
cd PaperBrain

2. Create a virtual environment:

# Using conda
conda create --name PaperBrain python=3.11
conda activate PaperBrain

# Using venv
python -m venv env
source env/bin/activate  # On Windows: .envScriptsactivate

3. Install dependencies:

pip install -r requirements.txt

4. Start required services:

docker-compose up -d

5. Pull required models:

# If you want other models
docker exec ollama ollama pull llama3.2:1b
docker exec -it ollama ollama pull mistral
docker exec -it ollama ollama pull nomic-embed-text

Usage

Paper Ingestion

python src/vector.py

• Place your markdown files in the markdowns/ directory
• System automatically processes and indexes papers
• Handles duplicate detection and tracking

Query Interface

python src/llmquery.py     #Run src/query.py to query qdrant database(without llm)

Available Commands

• quit or q: Exit the program
• analytics: Display system usage statistics
• clear: Reset paper history
• history: View recent questions and responses

Example Query Flow

> What are the main approaches for discovering governing equations from data?

The system will provide:
1. Main Answer: Comprehensive summary
2. Key Points: Important findings
3. Paper Citations: Relevant sources
4. Limitations: Gaps in current knowledge
5. Relevance Scores: Why papers were selected

Relevant Papers

? Project Structure

research-lens/
├── docker-compose.yml
├── requirements.txt
├── README.md
├── vector.py      # Paper ingestion and processing
├── llmquery.py          # Main Q&A interface
├── query.py          # To query qdrant databse without llm
├── markdowns/     # Paper storage directory
└── processed_papers.json # Paper tracking database

Configuration

Environment variables for system configuration:

QDRANT_HOST=localhost    # Qdrant server host
QDRANT_PORT=6333        # Qdrant server port
OLLAMA_HOST=localhost   # Ollama server host
OLLAMA_PORT=11434      # Ollama server port

Processing Pipeline

1. Paper Ingestion:

   • Reads markdown files from recommendations directory
   • Generates embeddings using Nomic Embed Text
   • Stores vectors and metadata in Qdrant
   • Tracks processed papers to avoid duplicates

2. Query Processing:

   • Converts user query to vector
   • Performs similarity search
   • Retrieves relevant papers
   • Generates structured LLM response

3. Response Generation:

   • Formats context for LLM
   • Generates structured response
   • Provides relevance explanations
   • Maintains conversation history

Future Roadmap

• Export functionality (PDF, markdown)
• Advanced paper filtering options
• Citation network visualization
• Multi-language support
• Batch processing capabilities
• API interface for integration
• Enhanced analytics dashboard
• Custom prompt templates

? Contributing

Contributions are welcome! Please:

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

? License

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

Acknowledgments

• Qdrant team for vector database
• Ollama project for LLM interface
• Nomic AI for embedding model
• LLaMA team for the base model
• The Markdowns were fetched using literatureSurvey

Citation

If you use this project in your research, please cite:

@software{PaperBrain_2024,
  author = {Ansh Kumar and Apoorva Gupta},
  title = {PaperBrain: Intelligent Research Paper Q&A System},
  year = {2024},
  url = {https://github.com/ansh-info/PaperBrain}
}