# Arbor

> Arbor is an AI-powered academic research workbench. Users enter a research question, and AI agents decompose it into focused inquiries, search academic papers across arXiv and Semantic Scholar, extract findings, and synthesize everything into a structured knowledge graph — all streamed in real-time.

Arbor is built for anyone with a research question, not just academics. The entire research pipeline runs automatically: question decomposition, parallel paper search, screening, finding extraction, and synthesis. Results are displayed as an interactive graph that users can explore, share, and export.

## Features

- [Research Pipeline](https://arborinquiries.com): Enter a question and watch AI agents build a knowledge graph in real-time
- [Paper Search](https://arborinquiries.com): Searches arXiv and Semantic Scholar for relevant academic papers
- [Knowledge Graph](https://arborinquiries.com): Interactive graph visualization showing questions, inquiries, papers, findings, and synthesis
- [Sharing](https://arborinquiries.com): Share research graphs via unique links
- [Export](https://arborinquiries.com): Export research findings as markdown

## How It Works

- A research question is decomposed into 4 focused inquiries by AI
- Each inquiry triggers parallel paper searches across arXiv and Semantic Scholar
- Papers are screened for relevance and findings are extracted
- All findings are synthesized into a structured summary
- The entire process streams in real-time via server-sent events, visualized as a growing graph

## Tech Stack

- Frontend: React, TypeScript, Vite, Tailwind CSS, React Flow
- Backend: FastAPI, Python, SQLite, SQLAlchemy
- LLM Providers: DeepSeek V3, Gemini 2.0 Flash, GPT-4o-mini (via OpenRouter)
- Hosting: Vercel (frontend), Railway (backend)

## Links

- [Website](https://arborinquiries.com)
- [Nur Opus](https://nuropus.com): Built by Nur Opus