# Code of Paper
> Paper-to-code search platform. Find GitHub implementations for research papers.

## What This Site Does
Code of Paper helps researchers and developers find code implementations of academic papers.
It indexes a growing collection of papers from arXiv and 7 major ML conferences (NeurIPS, CVPR, ICLR, ICML, ICCV, ECCV, WACV).
Papers are linked to GitHub repositories with quality scoring based on stars, recency, and official status.

## Website
- Homepage: https://codeofpaper.com
- Categories: https://codeofpaper.com/categories (70+ categories across 8 research areas)
- Conferences: https://codeofpaper.com/conferences
- Paper detail: https://codeofpaper.com/paper/{arxiv_id}

## API
- Base URL: https://api.codeofpaper.com
- Interactive docs: https://api.codeofpaper.com/docs
- OpenAPI spec: https://api.codeofpaper.com/openapi.json
- Authentication: GitHub OAuth (optional — most endpoints are public)

## Key API Endpoints (all public, no auth required)
- GET /papers/search?query={query}&limit=20 — Search papers by keyword
- GET /papers/{arxiv_id} — Get paper details (title, abstract, authors, categories)
- GET /papers/{arxiv_id}/repos — Get linked GitHub repositories for a paper
- GET /papers/{arxiv_id}/similar — Find semantically similar papers
- GET /trending/?sort=hot&limit=20 — Trending papers with code
- GET /categories/ — All categories grouped by research area
- GET /categories/{id}/papers — Papers in a specific category
- GET /conferences/ — All conference series with stats
- GET /conferences/{id} — Conference details and paper list
- GET /papers/search?query={query}&sort=has_code — Filter for papers with code only

## CLI Tool
- Install: pip install codeofpaper
- Entrypoints: both `codeofpaper` and `cop` work
- PyPI: https://pypi.org/project/codeofpaper/
- 20 commands, 6 output formats (table, json, csv, jsonl, brief, detail)
- Designed for scripting, pipelines, and LLM agent tool-calling

### Key CLI Commands
- cop search "world models" --limit 10 --format json — Search papers by keyword
- cop trending --sort hot --limit 20 — Get trending papers with code
- cop paper 2602.18998 — Get details for a specific paper
- cop paper 2602.18998 --repos — Get linked GitHub repos for a paper
- cop categories — List all categories
- cop conferences — List all conferences
- cop category cs.CV --limit 10 — Papers in a specific category
- cop search "transformers" --has-code --format csv — Filter for papers with code, CSV output

### Example Pipeline (daily digest of world model papers)
```bash
cop search "world models" --has-code --sort new --limit 20 --format json | jq '.[] | {title, arxiv_id, repos}'
```

## Coverage
- 70+ categories across 8 areas: Computer Vision, NLP, Multimodal, Learning Paradigms, Architectures, Theory, Trust & Safety, Applications
- 7 conference series: NeurIPS, CVPR, ICLR, ICML, ICCV, ECCV, WACV
- Repository quality signals: GitHub stars, official code flags, framework detection, license info

## Structured Data
- Paper pages include Schema.org ScholarlyArticle markup (JSON-LD)
- Conference pages include Schema.org Event markup (JSON-LD)
- Homepage includes WebSite + SearchAction markup for sitelinks search