Local-first reader that folds papers, web pages, files and AI replies into five levels of detail, from a one-screen map down to the full text.
aeroxy/paper-bro
Paste an arXiv paper link and get an interactive, step-by-step course generated from it, hosted on AERO.
- Stars
- 13
- Forks
- 1
- Open issues
- 4
- Last commit
- 4 days ago
What is it?
paper-bro turns an arXiv paper into an interactive course. You paste a paper's arXiv ID or URL, and a pipeline reads the text, builds an outline, writes lessons with embedded labs, checks them in a headless browser, and publishes the result. Each paper gets one course: the first request starts the build, and later visitors are sent to the same page.
The server is a single Rust binary (axum and tokio) with a Neon Postgres database and a React front end. Model calls go to any OpenAI-compatible endpoint, and each build stage is checkpointed so an interrupted job resumes. Progress streams to the paper page over SSE. Publishing uses AERO's MCP endpoint with a personal access token.
Lesson text is sanitized, and labs run in sandboxed iframes with an opaque origin, so model-written code cannot reach cookies, storage or the network of the host site. The project is intended for self-hosting behind a reverse proxy.
Key features
- 01Turns an arXiv paper ID or URL into a multi-lesson interactive course.
- 02Generates lessons in parallel and checks them with headless-browser QA before publishing.
- 03Resumes interrupted builds from stored checkpoints instead of regenerating everything.
- 04Streams build progress to the paper page in real time.
- 05Runs model-written labs in sandboxed iframes isolated from the host site.
- 06Works with any OpenAI-compatible model server.
Use cases
- Studying a research paper with guided lessons and quizzes instead of reading it cold.
- Hosting a self-run course generator for a team or class working through a set of papers.
- Reviewing how a sandboxed LLM content pipeline is structured in Rust.
Quick start
- 1
Copy the environment template and fill in the database, AERO token and LLM settings
cp .env.example .env - 2
Install the headless Chrome build from Chrome for Testing and point CHROME_PATH at it
- 3
Build the front end
(cd frontend && bun install && bun run build) - 4
Start the server
cargo run -- serve
Things to know
- Needs a Neon Postgres database, an AERO token and an OpenAI-compatible model API key.
- Each course build costs roughly $1 in model fees.
- Requires a headless Chromium; on Ubuntu 24.04 it needs an AppArmor profile to run.
In collections
Topics
FAQ
- What is paper-bro?
- Paste an arXiv paper link and get an interactive, step-by-step course generated from it, hosted on AERO. paper-bro turns an arXiv paper into an interactive course. You paste a paper's arXiv ID or URL, and a pipeline reads the text, builds an outline, writes lessons with embedded labs, checks them in a headless browser, and publishes the result. Each paper gets one course: the first request starts the build, and later visitors are sent to the same page.
- Is paper-bro free to use? What license does it have?
- paper-bro is open source under the MIT license. Check the license terms before using it in commercial products.
- Is paper-bro still maintained?
- Actively maintained. The last commit to paper-bro was on Nov 3, 2026.
- What can I use paper-bro for?
- Studying a research paper with guided lessons and quizzes instead of reading it cold. Hosting a self-run course generator for a team or class working through a set of papers. Reviewing how a sandboxed LLM content pipeline is structured in Rust.
- How do I install paper-bro?
- Copy the environment template and fill in the database, AERO token and LLM settings: cp .env.example .env Install the headless Chrome build from Chrome for Testing and point CHROME_PATH at it Build the front end: (cd frontend && bun install && bun run build) Start the server: cargo run -- serve
- Where can I get paper-bro?
- The source code and installation instructions are on GitHub: https://github.com/aeroxy/paper-bro
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Open-source platform that turns any topic or document into an interactive classroom with AI teachers, slides, quizzes and simulations.
An Agent Skill that turns a topic, text, web page, screenshot or PDF into a one-question-at-a-time Socratic tutoring session.
A curated, Markdown-based ML knowledge base of 923 papers, lectures and articles, plus a small NumPy-only custom classifier and regressor called APRL.
An interactive browser guide to the ESP32-S3 with 32 labs, live simulators, a pinout, a firmware code reader and a datasheet section.
An agent skill that turns a lesson claim into an interactive figure where learners predict, change a cause, and explain a new case.