ScholarAI — Academic Research Search Engine

AI-enhanced academic paper discovery with hybrid ranking
Powered by OpenAlex (260M+ scholarly works) · Scopus verification · Local AI summaries

---

Why ScholarAI?

Large language models like ChatGPT and Gemini often hallucinate citations — generating fake paper titles, non-existent DOIs, and fabricated authors. ScholarAI solves this by searching real academic databases directly.

	ChatGPT / Gemini	Google Scholar	ScholarAI
Real papers with DOIs	❌ Often hallucinated	✅	✅
Live citation counts	❌	✅	✅
AI-powered summaries	✅ (but fake sources)	❌	✅ (real source + AI)
Scopus indexed badge	❌	❌	✅
Citation export (BibTeX/APA)	❌ Unreliable	❌	✅ Formatted correctly
Excel bulk export	❌	❌	✅
Advanced filters	❌ Limited	❌ Basic	✅ Year, citations, type, author, OA
Hybrid ranking algorithm	❌	❌ Opaque	✅ Relevance 50% + Citations 30% + Recency 20%
Open source & self-hosted	❌	❌	✅

Features

• Hybrid Ranking — Papers scored by relevance (50%), citation count (30%), and recency (20%)
• AI Summaries — Local Llama/Ollama integration (free, private, no API key needed)
• Scopus Verification — Shows "Scopus Indexed" badge with direct Scopus links
• Advanced Filters — Year range, minimum citations, paper type, author name, open access only
• Citation Export — BibTeX and APA format, single paper or batch
• Excel Export — Download full search results as .xlsx
• Rate Limiting — Built-in abuse protection for production deployment
• Security Headers — HSTS, X-Frame-Options, XSS protection, CORS hardening

Architecture

┌─────────────┐     ┌──────────────────────┐     ┌─────────────────┐
│  Frontend   │────▶│   Backend (FastAPI)   │────▶│   OpenAlex API  │
│  Next.js    │     │                      │     │   260M+ papers  │
│  Tailwind   │     │  ┌────────────────┐  │     └─────────────────┘
│  shadcn/ui  │     │  │ Ollama (Llama) │  │     ┌─────────────────┐
└─────────────┘     │  │ AI Summaries   │  │────▶│   Scopus API    │
                    │  └────────────────┘  │     │   Verification  │
                    └──────────────────────┘     └─────────────────┘

Quick Start

Prerequisites

• Python 3.10+
• Node.js 18+
• Ollama (optional, for AI summaries) — ollama.com

1. Clone & Configure

git clone https://github.com/YOUR_USERNAME/scholar-ai.git
cd scholar-ai

# Create environment file
cp .env.example .env
# Edit .env with your settings

2. Backend

pip install -r requirements.txt
python main.py

API runs at http://localhost:9999 · Docs at http://localhost:9999/docs

3. Frontend

cd web
npm install
npm run dev

Frontend runs at http://localhost:3000

4. AI Summaries (Optional)

# Install Ollama from https://ollama.com, then:
ollama pull llama3.1

The backend auto-detects Ollama and enables AI summaries — no configuration needed.

Configuration

Copy .env.example to .env and customize:

Variable	Description	Default
APP_ENV	development or production	development
LLM_PROVIDER	auto, ollama, or openai	auto
OLLAMA_MODEL	Ollama model name	llama3.2
SCOPUS_API_KEY	Free key from dev.elsevier.com	—
OPENALEX_EMAIL	Email for faster API rate limits	—
CORS_ORIGINS	Allowed frontend origins	*

See .env.example for all options including rate limits and OpenAI fallback.

API Endpoints

Method	Endpoint	Description	Rate Limit
GET	/search	Search papers with hybrid ranking	30/min
GET	/scopus/check	Check Scopus indexing by DOI	20/min
POST	/summarize	AI paper summary (Ollama/OpenAI)	10/min
POST	/cite	Generate BibTeX or APA citation	30/min
POST	/cite/batch	Batch citation export	5/min
GET	/export	Export results to Excel (.xlsx)	5/min
GET	/trending	Publication trends by field	30/min
GET	/health	System status & feature availability	60/min

Tech Stack

• Backend: Python 3, FastAPI, slowapi (rate limiting)
• Frontend: Next.js, TypeScript, Tailwind CSS, shadcn/ui, Lucide Icons
• AI: Ollama + Llama 3 (local, free) · OpenAI (optional fallback)
• Data: OpenAlex API (260M+ works) · Scopus API (indexing verification)

Production Deployment

For production, set in .env:

APP_ENV=production
CORS_ORIGINS=https://your-domain.com

Production mode enables:

• Security headers (HSTS, X-Frame-Options, CSP)
• API docs hidden
• Binds to 0.0.0.0 for reverse proxy / tunnel access
• Structured logging with request timing

Data Sources & Credibility

All paper metadata comes from OpenAlex — a free, open catalog of 260M+ scholarly works maintained by a nonprofit. Data is sourced from Crossref, PubMed, DOAJ, and publisher feeds. Every paper has a verifiable DOI.

AI summaries are the only generated content and are clearly labeled. Paper metadata is never hallucinated.

Citation: Priem, J., Piwowar, H., & Orr, R. (2022). OpenAlex: A fully-open index of scholarly works, authors, venues, institutions, and concepts. ArXiv. https://arxiv.org/abs/2205.01833

License

MIT — free for academic and commercial use.