rajni-research-assistant: An Academic Paper Automatic Analysis Tool Based on Multi-Agent Architecture

rajni-research-assistant is an open-source academic research assistant that adopts a multi-agent architecture, combining technologies like Streamlit, LangChain, ChromaDB, and Gemini/Groq to achieve automated paper analysis, knowledge extraction, slide generation, and defense question prediction. It is suitable for researchers, graduate students, and educators, aiming to simplify academic workflows and improve efficiency.

The original author/maintainer is rajnitiwari28, and the project is open-sourced on GitHub (link: https://github.com/rajnitiwari28/rajni-research-assistant), released on June 16, 2026. This tool is specifically designed for academic researchers, capable of converting PDF papers into structured knowledge products, supporting scenarios such as quick literature comprehension, defense preparation, and presentation creation. It is suitable for researchers, graduate students, and educators who need to process large volumes of literature.

The system includes four agents: 1. Abstract Generation Agent: Extracts core logic such as research motivation, methods, and results; 2. Concept Extraction Agent: Identifies key terms and technical concepts; 3. Slide Generation Agent: Generates downloadable PPTs via python-pptx; 4. Defense Prediction Agent: Predicts defense questions and provides reference answers.

The technology stack includes: Frontend Streamlit, LLM (Gemini default / Groq optional), vector database ChromaDB (in-memory mode), text embedding Sentence-Transformers (MiniLM), process orchestration LangChain, and document processing PyPDF. All agents share an LLM factory, and Gemini/Groq can be switched via environment variables, offering high flexibility.

The flow is: PDF → PyPDF loading → Text segmentation → MiniLM embedding → ChromaDB storage → Processing by various agents (abstract/concept extraction, slide generation, defense prediction). Slide generation outputs PPTs, defense prediction outputs JSON-format Q&A, and collaboration between agents is efficient.

Local Run: Clone the repository → Create a virtual environment → Install dependencies → Configure API keys → Run streamlit run app.py. Free Deployment: Push code to GitHub → Connect to Streamlit Community Cloud → Add API keys in Secrets for zero-cost sharing.

Developers can customize: Modify prompt templates in the agents directory to adjust outputs; Add new agents (e.g., review agent); Enable ChromaDB persistent storage; Integrate other LLM providers in llm_factory.py.

Project Value: Lowers the threshold for literature reading, improves defense preparation efficiency, simplifies academic sharing, and serves as a learning example for LangChain/RAG. Summary: This tool is a well-thought-out multi-agent system, providing a practical assistant for researchers and a reference architecture for developers, making it worth attention and trial.