Analysis updated 2026-06-24
Publish preprints as federated AT Protocol records that live in the author's own PDS
Run a self-hosted open preprint server with OAI-PMH for library indexers
Show ORCID, citation, and integrity badges on arxiv.org pages via the browser extension
Declare AI use on a paper with a structured disclosure record of none, assistant, coauthor, or primary
| davidichalfyorov-wq/openxiv | kentjuno/kjaudiobook-v1 | thedecipherist/markdownai | |
|---|---|---|---|
| Stars | 33 | 33 | 33 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | hard | hard | easy |
| Complexity | 5/5 | 4/5 | 3/5 |
| Audience | researcher | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Needs Node 22, PostgreSQL with pgvector, Redis, S3 storage, and Docker-sandboxed Tectonic plus GROBID for the LaTeX pipeline.
OpenXiv is an open preprint server for scientific papers built on top of the AT Protocol, the same decentralized network that powers Bluesky. Papers, threads, endorsements, AI-use disclosures, and reviews are not stored as private database rows, they are real records under the namespace app.openxiv.* that live in each author's PDS, the personal data store provided by Bluesky or by the author themselves. From there they federate out into the wider AT Protocol network. The live site is openxiv.net, the project has an ISSN and a Wikidata entry, and it is indexed by science indexers like BASE, CORE, and the Bluesky directory of custom feeds. The project is explicitly open to independent researchers without institutional backing. You sign in with ORCID, Google, or your Bluesky did:plc identity. There is no endorsement gate and no affiliation filter. AI use is allowed but has to be declared with a structured disclosure value, one of none, assistant, coauthor, or primary, recorded as an app.openxiv.disclosure record. If a submission fails policy checks, the server returns a public refusal packet that names the failure mode and lets the work come back as a revision, rather than banning the author. The README lists a long table of what is already live: stable paper ids of the form subject.year.NNNNN, profile pages, a six-stage submission saga with per-stage retry, a paper page at /abs/{id} with PDF, HTML, three explainer tiers, and a saga timeline, a Trust Passport that exposes lanes for identity, provenance, citations, integrity, and external attestations, an OAI-PMH 2.0 endpoint for library indexers, six custom Bluesky feeds, refusal packets, a browser extension that overlays trust badges on arxiv.org pages, and pre-registration records. The stack is TypeScript on Node 22. Fastify 5 serves the API with zod-typed routes, PostgreSQL 16 with pgvector and Drizzle ORM holds the database, BullMQ on Redis 7 runs the heavy jobs, and S3-compatible storage holds the files. Tectonic compiles LaTeX to PDF in a sandboxed Docker container, latexml produces HTML, GROBID pulls out metadata, and DeepSeek V4 Flash with Gemini embeddings handles explainer text and similarity. Astro 5 with React 19 islands renders the web pages. The project is licensed AGPL-3.0-or-later.
An open preprint server built on the AT Protocol where papers, reviews, and AI-use disclosures live as federated records in each author's Bluesky PDS rather than a private database.
Mainly TypeScript. The stack also includes TypeScript, Fastify, PostgreSQL.
AGPL-3.0-or-later: free to use and modify, but any networked service built on it must also be released under AGPL.
Setup difficulty is rated hard, with roughly 1day+ to a first successful run.
Mainly researcher.
This repo across BitVibe Labs
Verify against the repo before relying on details.