LitMine

AI-powered literature mining and research analytics platform for researchers across all disciplines.

中文 | 日本語 | English

---

What is LitMine

LitMine is an AI-powered platform that helps researchers complete literature surveys faster. It does not rely on any preset domain vocabulary — whether you work in synthetic biology, materials science, or social sciences, LitMine extracts keywords from your paper corpus automatically, identifies research hotspots, finds gaps, and generates structured literature reviews.

The traditional literature survey workflow — searching PubMed / Google Scholar paper by paper → manually organizing notes → judging hotspots and gaps yourself → writing the review — is compressed by LitMine into minutes.

---

Features

• Paper Discovery — Search across arXiv, PubMed, bioRxiv, and OpenAlex simultaneously with automatic deduplication and metadata backfill
• Paper Import — Upload PDFs (auto-parsed via GROBID for title, authors, abstract, DOI), paste DOI links, or import BibTeX/RIS files
• Multi-language Translation — Translate paper titles and abstracts into Chinese, Japanese, Korean, Spanish, or Italian. Results cached to database — no repeated API calls
• AI Deep Analysis — Seven-dimension analysis: summary, core contribution, methodology, key results, key takeaways, limitations, and deep insights
• Paper Comparison — Select multiple papers and let AI compare methodology, findings, strengths, limitations, and research gaps
• Chat with Paper — Ask questions about any paper and get contextual answers
• Research Hotspots & Gaps — Dynamic TF-IDF term extraction + trend slope sorting. Works for any research direction — no hardcoded vocabulary
• Literature Review Generation — 5-section structured report with three-line academic tables and a narrative introduction section ready for your paper
• Full-platform Multi-language — Translation, analysis, comparison, and reports all support six languages (EN/ZH/JA/KO/ES/IT)
• 3-container Deployment — PostgreSQL + Redis + GROBID, docker compose up -d to start

---

Screenshots

---

Quick Start

Prerequisites

• Docker Desktop
• Node.js 18+
• Python 3.11+
• An LLM API key (DeepSeek, OpenAI, Gemini, Anthropic, OpenRouter, Qwen, or Kimi — see LLM Settings in-app)

1. Start Services

docker compose up -d

Starts PostgreSQL (with pgvector extension), Redis, and GROBID.

2. Configure & Run Backend

cd apps/api
cp .env.template .env       # edit .env — (optional) set a server-default LLM API key
pip install -e .
alembic upgrade head
uvicorn app.main:app --host 0.0.0.0 --port 8000 --reload

3. Start Frontend

cd apps/web
npm install
npx next dev --port 3000

4. Open

Visit http://localhost:3000 → register an account → create your first project → Discover papers.

---

Architecture

litmine/
├── apps/
│   ├── api/          # FastAPI backend (Python 3.11+)
│   └── web/          # Next.js frontend (TypeScript)
├── docker-compose.yml
├── README.md
├── README_CN.md
├── ROADMAP.md
└── scripts/

Tech Stack

Layer	Technology
Backend	FastAPI + SQLAlchemy + asyncpg
Frontend	Next.js 16 (App Router) + TypeScript + Tailwind CSS
Database	PostgreSQL 16 + pgvector (semantic search)
Cache	Redis
PDF Parsing	GROBID 0.8.1 (TEI XML extraction)
LLM	DeepSeek (chat + translation + entity extraction)
Infra	3 Docker containers: PostgreSQL, Redis, GROBID

---

Key Design Decisions

Decision	Rationale
No Neo4j	Hotspot + Gap charts only; PostgreSQL GROUP BY is faster and lighter
No message queue	Single-user context — report generation runs inline; fewer moving parts
No object storage	PDFs stored on local filesystem; no external service dependency
Dynamic vocabulary	TF-IDF from user's paper corpus; cross-disciplinary seed words for cold start
LLM server-side proxy + multi-provider	7 providers supported (DeepSeek, OpenAI, Gemini, Anthropic, OpenRouter, Qwen, Kimi). User API keys stored locally in browser, all calls routed through backend proxy

---

Development

# Backend
cd apps/api && uvicorn app.main:app --reload

# Frontend
cd apps/web && npx next dev

---

Roadmap

See ROADMAP.md for the complete product roadmap and current status.

---

Buy Me a Coffee

If LitMine helps your research, consider buying the author a coffee!

_{WeChat                                                      Alipay}

Contact

• Email: zming19861028@126.com / zming19861028@hotmail.com
• WeChat: zminlibra
• GitHub: zminlibra

Acknowledgments

LitMine is built on the shoulders of many great open-source projects:

• GROBID — PDF to structured TEI XML extraction
• FastAPI — Python web framework
• Next.js — React framework
• Tailwind CSS — Utility-first CSS
• PostgreSQL + pgvector — Database & vector search
• Recharts — Charting library
• DeepSeek — LLM API
• All early-stage testers and colleagues who provided critical feedback during Phase 1–4

License

MIT © 2025–2026