resynth 0.1.2

Research Paper Synthesis Agent - Fetch, process, and query academic papers with AI-powered answers

ReSynth - Research Paper Synthesis Agent

ReSynth is an intelligent agent that fetches research papers, processes them through advanced chunking and embedding, and answers queries with proper academic citations. Perfect for researchers, students, and anyone working with academic literature.

Features

• Multi-Source Paper Fetching: Retrieve papers from arXiv, PubMed, and more
• Intelligent Processing: Advanced text chunking with semantic boundaries
• Vector Storage: Efficient storage and retrieval with ChromaDB
• AI-Powered Answers: Synthesize responses using OpenAI or local models
• Citation Management: Automatic citation generation in multiple styles (APA, MLA, numeric)
• Multiple Interfaces: Web UI, REST API, and command-line interface
• Quality Metrics: Retrieval quality validation and confidence scoring

Quick Start

Installation

# Install from PyPI
pip install resynth

# Or install with development dependencies
pip install resynth[dev]

Basic Usage

import resynth

# Initialize the agent
agent = resynth.ReSynthAgent()

# Search and process papers
papers = agent.search_and_process(
    query="machine learning interpretability",
    source="arxiv",
    max_papers=5
)

# Query the processed papers
answer = agent.query(
    "What are the main challenges in deep learning interpretability?",
    citation_style="apa"
)

print(answer.answer)
print(answer.bibliography)

Command Line Interface

# Search and process papers
resynth --search "transformer architectures" --max-papers 5

# Query processed papers
resynth --query "How do attention mechanisms work?" --citation-style numeric

# Show system statistics
resynth --stats

Web Interface

# Start the web interface
resynth-web

# Or use streamlit directly
streamlit run resynth.web

API Server

# Start the API server
resynth-api

# Then access the API at http://localhost:8000
# Interactive docs at http://localhost:8000/docs

Requirements

• Python 3.8+
• Optional: OpenAI API key for enhanced answer synthesis
• Optional: spaCy model (python -m spacy download en_core_web_sm)

Configuration

Create a .env file:

cp .env.example .env
# Edit .env with your configuration

Key configuration options:

# OpenAI API Key (optional, for enhanced synthesis)
OPENAI_API_KEY=your_openai_api_key_here

# Vector database settings
CHROMA_PERSIST_DIRECTORY=./chroma_db
CHROMA_COLLECTION_NAME=research_papers

# Retrieval settings
TOP_K_RETRIEVAL=5
SIMILARITY_THRESHOLD=0.7

# Chunking settings
CHUNK_SIZE=1000
CHUNK_OVERLAP=200

Usage Examples

Research Paper Analysis

import resynth

# Initialize agent
agent = resynth.ReSynthAgent()

# Process recent papers on a topic
agent.search_and_process(
    query="large language model alignment",
    source="arxiv",
    max_papers=10,
    fetch_content=True
)

# Get comprehensive analysis
answer = agent.query(
    "What are the main approaches to LLM alignment?",
    citation_style="author_date"
)

print(f"Answer: {answer.answer}")
print(f"Confidence: {answer.confidence_score:.2f}")
print(f"Sources: {len(answer.source_chunks)} papers")

Comparison Study

# Compare different methodologies
answer = agent.query(
    "Compare transformer architectures: BERT vs GPT vs T5",
    citation_style="apa"
)

# The answer will include:
# - Detailed comparison of architectures
# - Proper citations for each model
# - Confidence assessment
# - Source references

Literature Review

# Systematic literature review
agent.search_and_process(
    query="climate change machine learning applications",
    source="both",  # arxiv and pubmed
    max_papers=20
)

# Get overview
overview = agent.query(
    "What are the main applications of ML in climate research?",
    citation_style="numeric"
)

# Get specific methodology info
methods = agent.query(
    "What machine learning methods are most commonly used?",
    citation_style="mla"
)

Architecture

ReSynth/
├── Paper Fetchers     # arXiv, PubMed integration
├── Text Processors    # Chunking, cleaning, preprocessing
├── Embedding Engine  # Vector generation and storage
├── Retrieval System   # Query processing and similarity search
├── Answer Synthesizer # AI-powered response generation
└── Citation Manager   # Automatic citation formatting

Citation Styles

ReSynth supports multiple citation formats:

• APA: (Smith, 2023)
• MLA: Smith, John. "Title." Journal (2023)
• Numeric: [1], [2], [3]
• Author-Date: (Smith, 2023)

Quality Metrics

Every answer includes quality assessment:

• Confidence Score: 0.0-1.0 based on source quality
• Retrieval Quality: Validation of search results
• Source Diversity: Number of unique papers referenced
• Similarity Metrics: Average and minimum similarity scores

API Endpoints

When running the API server:

# Process papers
POST /process
{
  "query": "machine learning interpretability",
  "source": "arxiv",
  "max_papers": 5
}

# Query papers
POST /query
{
  "query": "What are interpretability methods?",
  "citation_style": "apa",
  "top_k": 5
}

# Get statistics
GET /stats

# List papers
GET /papers

Testing

# Run all tests
pytest

# Run with coverage
pytest --cov=resynth --cov-report=html

# Run specific test categories
pytest -m unit      # Unit tests only
pytest -m integration  # Integration tests only

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

Development Setup

# Clone the repository
git clone https://github.com/resynth-ai/resynth.git
cd resynth

# Install development dependencies
pip install -e ".[dev]"

# Set up pre-commit hooks
pre-commit install

# Run tests
make test

Documentation

• Full Documentation
• API Reference
• Examples

Roadmap

• Support for more paper sources (IEEE Xplore, Google Scholar)
• Advanced query expansion with semantic search
• Paper summarization and key point extraction
• Collaborative filtering and recommendation
• Export to various formats (LaTeX, Word, Markdown)
• Integration with reference managers (Zotero, Mendeley)

License

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

Acknowledgments

• Built with FastAPI, Streamlit, and ChromaDB
• Paper fetching powered by arXiv and PubMed
• Embeddings from OpenAI and Hugging Face
• Citation formatting inspired by academic standards

Support

• Report Issues
• Discussions
• Email

---

If you find ReSynth useful, please give us a star on GitHub!