Reesarch — Research Paper Question Answering

A lightweight, reproducible Retrieval-Augmented Generation (RAG) pipeline for answering questions about research papers.

Highlights

• Designed for accuracy and efficiency: small, task-specific models and two-stage retrieval.
• Grounded answers with source attribution and evaluation utilities.

Quick Start

• Clone and install:

git clone https://github.com/Priyans00/Reesarch
cd reesarch
python -m venv venv
venv\Scripts\activate    # Windows
pip install -r requirements.txt

• Run the web UI:

streamlit run app.py

Open http://localhost:8501 in your browser.

Programmatic use

from src.pipeline import create_pipeline
pipeline = create_pipeline()
pipeline.add_document("path/to/paper.pdf")
result = pipeline.query("What is the main contribution?")
print(result.answer)
print("Sources:", len(result.sources))

Repository layout

app.py
config.py
requirements.txt
data/         # uploads, processed artifacts, vector index
src/          # core modules: pdf_processor, chunker, embedder, vector_store, retriever, generator, pipeline

Configuration

• See config.py for the main runtime options (models, retrieval sizes, chunking).

Recommended sensible defaults are included; modify values only if you understand the downstream effects.

Design notes

• Embeddings: SPECTER (scientific-text tuned) for document similarity.
• Retrieval: dense vector search followed by a cross-encoder reranker for precision.
• Generation: constrained, low-temperature decoding with explicit grounding to retrieved context.

Troubleshooting

• Out of memory: reduce embedding batch sizes or CHUNK_SIZE; fall back to CPU.
• Slow: enable GPU, lower TOP_K_RETRIEVAL, or pre-load models.

License

• MIT

Acknowledgments

• SentenceTransformers, SPECTER, FAISS, and the Qwen model authors.