PaperReview

Jiphyeonjeon (집현전)

AI-Powered Academic Research Platform

Search. Review. Learn. — One platform for your entire research workflow.

[](https://jiphyeonjeon.kr) [](./LICENSE) [](https://python.org) [](https://react.dev)

---

---

Why Jiphyeonjeon?

Researchers spend hours jumping between Google Scholar, arXiv, PDF readers, and note apps. Jiphyeonjeon consolidates the entire workflow — search, review, annotate, learn — into a single platform powered by multi-agent AI.

• 6 academic databases searched in parallel (arXiv, Google Scholar, OpenAlex, DBLP, Connected Papers, Semantic Scholar)
• Multi-agent review pipeline with fact verification against source texts
• Conference-quality poster generation from review reports (NeurIPS / ICML / CVPR)
• Personalized learning curricula generated from your bookmarked papers
• Zero external DB required — JSON-file-based storage, deploy anywhere

> Named after the Jiphyeonjeon (집현전), the Hall of Worthies from the Joseon Dynasty — a royal research institute where scholars gathered to study and create knowledge.

---

Features

Research & Discovery

| Feature | What it does | How it works | |---------|-------------|--------------| | Multi-Mode Search | Find papers across 6 databases in one query | 3 modes: Basic (keyword), Smart (LLM query expansion), Deep (ArxivQA ReAct agent with rubric evaluation) | | Further Reading | Explore citation networks up to 3 levels deep | Semantic Scholar API — references, cited-by, influence scoring | | Knowledge Graph | Build and query entity-relationship graphs from papers | Custom LightRAG with 5 retrieval modes (naive, local, global, hybrid, mix) |

Analysis & Review

| Feature | What it does | How it works | |---------|-------------|--------------| | Deep Review | Generate systematic review reports | Multi-agent pipeline (LangGraph) — quality validation + fact verification. Fast / Deep modes | | Paper Review | Analyze individual papers in detail | Structured review with PDF highlight extraction + inline math explanation | | Notes & Highlights | Annotate across 6 categories | AI-generated + manual highlights, memos, BibTeX/Markdown export | | Chat with Papers | Q&A over your research collection | Streaming answers using reports + highlights + knowledge graph context |

Creation & Learning

| Feature | What it does | How it works | |---------|-------------|--------------| | Academic Poster | Generate conference-style posters | Paper2Poster binary-tree layout, NeurIPS/ICML/CVPR templates, auto SVG diagrams | | Auto Figure | Convert methodology descriptions to diagrams | PaperBanana SVG generation with Gemini fallback | | Learning Curriculum | AI-generated learning paths from bookmarks | Per-module progress tracking, fork & share via public links | | Share | Share bookmarks and curricula externally | Read-only links with configurable expiration |

---

Architecture

User Query
  │
  ├─ SearchAgent ─── arXiv / Scholar / OpenAlex / DBLP / Connected Papers / S2
  │                    ↓
  │              BM25 + FAISS + LLM Rerank → Deduplicated results
  │
  ├─ DeepAgent ──── Multi-agent review pipeline (LangGraph)
  │                    ↓
  │              Quality validation → Fact verification → Review report
  │
  ├─ QueryAgent ─── Query analysis, diversification, rubric evaluation
  │
  └─ GraphRAG ──── Entity extraction → Knowledge graph → 5-mode retrieval

Multi-Agent Pipeline

The review pipeline orchestrates specialized LLM agents through LangGraph:

1. Query Analysis — Decompose research questions, diversify search terms
2. Parallel Search — Hit 6 databases concurrently, deduplicate via DOI/title similarity
3. Hybrid Ranking — BM25 (lexical) + FAISS (semantic) + LLM rerank (relevance judge)
4. Deep Review — Multi-pass analysis with section-level fact verification
5. Post-Processing — Poster generation, figure synthesis, curriculum creation

---

Tech Stack

| Layer | Technologies | |-------|-------------| | Frontend | React 19, TypeScript, Vite 7, React Router, Plotly.js, dnd-kit | | Backend | FastAPI, Python 3.12, Slowapi (rate limiting), JWT + bcrypt | | AI / LLM | GPT-4.1, GPT-4o-mini, Google Gemini, text-embedding-3-small, sentence-transformers | | Diagrams & Posters | PaperBanana (SVG generation), Playwright (HTML → PDF/PNG export) | | PDF Processing | PyMuPDF, pdfplumber, PyPDF2 | | Search & Retrieval | BM25 Okapi, FAISS, NetworkX, LangChain 0.3, LangGraph 0.2 | | External APIs | arXiv, Google Scholar, OpenAlex, DBLP, Connected Papers, Semantic Scholar | | Infrastructure | AWS EC2, Nginx, Let's Encrypt, Docker |

---

Quick Start

Option 1: Docker (recommended)

git clone https://github.com/KimJiSeong1994/PaperReview.git
cd PaperReview
cp .env.example .env      # configure your keys
docker compose up -d       # → http://localhost:8000

Option 2: Manual

# Backend
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python api_server.py                      # → http://localhost:8000

# Frontend
cd web-ui && npm install && npm run dev   # → http://localhost:5173

Environment Variables

| Variable | Required | Description | |----------|----------|-------------| | OPENAI_API_KEY | Yes | OpenAI API key for GPT-4.1, embeddings | | JWT_SECRET | Yes | JWT signing secret (min 32 chars) | | S2_API_KEY | No | Semantic Scholar API key (relaxes rate limits) | | GOOGLE_API_KEY | No | Google Gemini API key (poster/diagram generation) | | CORS_ORIGINS | No | Allowed origins, comma-separated | | API_AUTH_KEY | No | Optional API-level auth key | | REQUEST_TIMEOUT | No | Slow-log threshold in seconds (default: 120) |

---

API Overview

All endpoints are prefixed with /api. Authentication uses JWT Bearer tokens. Interactive docs: jiphyeonjeon.kr/docs (Swagger UI)

| Group | Key Endpoints | Description | |-------|--------------|-------------| | Auth | POST /register, /login | JWT authentication | | Search | POST /search, /smart-search, /deep-search | Three search modes | | Papers | POST /save, /extract-texts, /enrich-papers | Paper storage & enrichment | | Reviews | POST /deep-review, GET /deep-review/status/{id} | Async deep review with polling | | Paper Reviews | POST /bookmarks/{id}/papers/{idx}/review, /math-explain | Per-paper review & math explanation | | Bookmarks | POST /bookmarks, /bookmarks/{id}/auto-highlight | Bookmark management & AI highlights | | Curriculum | POST /curricula/generate, /curricula/{id}/fork | Learning path generation | | Chat | POST /chat | Streaming Q&A | | Knowledge Graph | POST /light-rag/build, /light-rag/query | LightRAG build & query | | Exploration | POST /bookmarks/{id}/citation-tree | Citation tree traversal | | Auto Figure | POST /autofigure/method-to-svg | SVG diagram generation | | PDF | GET /pdf/proxy, /pdf/resolve | PDF proxy & URL resolution | | Share | POST /bookmarks/{id}/share, GET /shared/{token} | Share link management | | Admin | GET /admin/dashboard, /admin/users | System management |

---

Project Layout

api_server.py          FastAPI entrypoint — middleware, router registration
routers/               14 API routers
  ├── auth.py            Register / login / JWT verify
  ├── search.py          Basic, Smart, Deep search (ArxivQA)
  ├── papers.py          Paper CRUD, references, code repos, graph data
  ├── reviews.py         Deep review pipeline + poster visualization
  ├── paper_reviews.py   Individual paper review, PDF highlights, math explain
  ├── bookmarks.py       Bookmark CRUD, auto-highlight, bulk ops
  ├── curriculum.py      Learning curriculum generate / fork / share
  ├── chat.py            Streaming Q&A over bookmarked papers
  ├── lightrag.py        Knowledge graph build / query / status
  ├── exploration.py     Citation tree (Semantic Scholar)
  ├── autofigure.py      PaperBanana SVG diagram generation
  ├── pdf_proxy.py       PDF proxy, URL resolve, batch resolve
  ├── share.py           Read-only share links with expiration
  └── admin.py           Dashboard, user/paper/bookmark management
app/                   Agent modules
  ├── SearchAgent/       Multi-source parallel search
  ├── QueryAgent/        Query analysis, diversification, rubric evaluation
  ├── DeepAgent/         Multi-agent review pipeline (LangGraph)
  └── GraphRAG/          Graph-based retrieval-augmented generation
src/                   Core libraries
  ├── collector/         Paper collection from external APIs
  ├── graph/             Citation graph construction (NetworkX)
  ├── graph_rag/         Hybrid ranking (BM25 + FAISS + LLM rerank)
  ├── light_rag/         Custom LightRAG implementation
  └── utils/             Shared utilities
web-ui/                React frontend
  ├── src/components/    Page components (MyPage, Curriculum, Admin, ...)
  │   ├── mypage/          Bookmark sidebar, chat, paper viewer, report viewer
  │   └── curriculum/      Course sidebar, module view, detail panel
  ├── src/hooks/         Custom hooks (useDeepReview, useCurriculum, ...)
  └── src/api/           API client
data/                  JSON storage + FAISS indices + caches

---

Contributing

Contributions welcome! Open an issue or submit a PR.

For coding conventions, see [.claude/rules/](.claude/rules/) — Python (PEP 8, type hints), TypeScript (strict mode, interface-first), API design patterns.

---

References

• Robertson, S. E. et al. (1995). Okapi at TREC-3. NIST Special Publication, 500-225.
• Johnson, J. et al. (2019). Billion-scale similarity search with GPUs. IEEE Trans. Big Data, 7(3).
• Hagberg, A. A. et al. (2008). Exploring network structure using NetworkX. SciPy, 11-15.
• Guo, Z. et al. (2024). LightRAG: Simple and Fast Retrieval-Augmented Generation. arXiv:2410.05779.
• Lewis, P. et al. (2020). Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. NeurIPS, 33.
• Mao, K. et al. (2024). ArxivQA: A Dataset for Paper Retrieval Agent Evaluation. arXiv.

---

License

[Apache License 2.0](./LICENSE)

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: KimJiSeong1994
• Source: KimJiSeong1994/PaperReview
• License: Apache-2.0
• Homepage: https://jiphyeonjeon.kr

Install and usage instructions live in the source repository linked above.