# Tomesphere

> An interactive research paper map of every paper on arXiv. 3M+ papers, organized into 500 semantic fields via SPECTER2 embeddings and Louvain community detection. Each paper has an editorial brief (TL;DR, contribution, findings, code, models, citation network, semantic neighbors) and each field has a guided briefing (canon, what's rising, 3-paper reading path).

Built for working researchers entering or browsing fields. Open data sources: arXiv metadata, OpenAlex citations, Hugging Face models/datasets, GitHub repos, OpenReview peer reviews. Tomesphere is the canonical brand; paperverse.science is a legacy redirect domain.

## Primary pages

- [Home](https://tomesphere.com/): landing page with search, atlas entry, extension link
- [Fields directory](https://tomesphere.com/clusters): all 500 research subfields as searchable cards
- [Atlas](https://tomesphere.com/atlas): interactive 500K–3M paper map with WebGL rendering, color modes, time scrub
- [Paper page](https://tomesphere.com/paper/{arxiv_id}): per-paper page with TL;DR, contribution, findings, abstract, peer reviews, code & models, videos, citation network, semantic graph, references, taxonomy. Example: https://tomesphere.com/paper/1502.03167
- [Field briefing](https://tomesphere.com/cluster/{id}): per-cluster editorial briefing with subfield name, summary, key methods, open questions, 3-paper reading path, canon, surveys, year-by-year. Example: https://tomesphere.com/cluster/104
- [Browser extension](https://tomesphere.com/extension/welcome): Chrome extension that adds a side panel to arxiv.org pages with the same TLDR + neighbors + code data
- [Privacy policy](https://tomesphere.com/privacy)

## API

The site exposes a small JSON read API. All endpoints are public, rate-limited per IP.

- [GET /api/paper/by-arxiv/{arxiv_id}](https://tomesphere.com/api/paper/by-arxiv/1502.03167): full paper record (title, authors, abstract, tldr, contribution, findings, code, models, datasets, videos, references, similar)
- [GET /api/paper/{arxiv_id}/citations](https://tomesphere.com/api/paper/1502.03167/citations): focal paper + references (outbound) + citations (inbound) for the citation network graph
- [GET /api/paper/{arxiv_id}/semantic?mode=influential|recent|nearest|hidden&limit=40](https://tomesphere.com/api/paper/1502.03167/semantic): SPECTER2-nearest neighbors with mode filtering
- [GET /api/search?q=…](https://tomesphere.com/api/search): paper title + author search
- [GET /api/map/meta?indices=…&tier=top500k|full](https://tomesphere.com/api/map/meta): bulk metadata for atlas point indices

## Data sources

- arXiv (paper metadata + abstracts)
- OpenAlex (citations, references, venues)
- Semantic Scholar fallback for missing citations
- Hugging Face hub-stats (papers ↔ models/datasets)
- GitHub (from LLM extraction + hub-stats + per-paper API)
- OpenReview (peer reviews — CVPR/ECCV/ICCV/ICLR/NeurIPS/AAAI etc.)
- YouTube + SlidesLive + Underline (talks/videos)

## Permissions for crawlers and AI agents

Tomesphere allows crawling by all major AI bots (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, etc.) and search engines. See https://tomesphere.com/robots.txt for the full allowlist. API endpoints under /api/ are disallowed for crawl efficiency — use the JSON endpoints above instead.

## Contact

- Support: social@tomesphere.com
- DNS: Cloudflare
- Source: paper data pipeline + clustering + Next.js app