Repolode
Submit
evanhu1

evanhu1/talk2arxiv

Swap arxiv.org for talk2arxiv.org in any paper link to read the paper and chat with an AI that has read its full text.

View on GitHubWebsite
Stars
531
Forks
31
Open issues
3
Last commit
last month

What is it?

Talk2Arxiv is a web app that lets you read an arXiv paper and ask an AI questions about it. Changing the domain in any arXiv link to talk2arxiv.org opens the paper with a table of contents, a reading-progress indicator, and a chat panel. The same pattern works for bioRxiv papers on talk2biorxiv.org.

HTML versions of papers are loaded from arXiv's HTML5 rendering, with ar5iv as a fallback, and equations are shown as native MathML. Papers without an HTML version open as PDFs rendered in the browser, and the model reads the PDF directly, including figures. Selecting a passage lets you send it to the chat with its LaTeX equations, and clicking a bibliography citation shows a preview of the referenced paper.

The model receives the whole paper text in its context rather than using chunking, embeddings, or a vector database. Requests go to a Cloudflare Worker that calls GPT-6 Luna through OpenRouter. Answers are cached for identical questions on the same paper, and chat history is stored in the browser's localStorage.

Key features

  • 01Reads arXiv papers in HTML form with a table of contents and reading progress.
  • 02Falls back to a PDF viewer for papers without HTML versions.
  • 03Lets you select a passage and send it to the chat with Ask AI or Explain.
  • 04Previews cited references and can add them to the chat context.
  • 05Puts the full paper text into the model context without chunking or embeddings.
  • 06Stores chat history locally in the browser.

Use cases

  • Getting a quick summary of a new arXiv paper before reading it closely.
  • Asking clarifying questions about a method, equation, or limitation in a paper.
  • Checking a paper's claims against its evidence while reading.
  • Exploring the work of a paper's authors and the papers it cites.

Quick start

  1. 1

    Install dependencies

    yarn
  2. 2

    Copy the example env file and add your OpenRouter key

    cp .dev.vars.example .dev.vars
  3. 3

    Start the dev server

    yarn dev

Things to know

  • Chat requires an OpenRouter API key, which may incur usage costs.
  • Papers longer than about 800K tokens get an error, and papers arXiv has not rendered as HTML cannot be read.

In collections

PDF, OCR & document toolsAI chatbots & chat clients

Topics

#arxiv#llm#research#rag#pdf#academic-papers#chatbot#gpt#open-source

FAQ

What is talk2arxiv?
Swap arxiv.org for talk2arxiv.org in any paper link to read the paper and chat with an AI that has read its full text. Talk2Arxiv is a web app that lets you read an arXiv paper and ask an AI questions about it. Changing the domain in any arXiv link to talk2arxiv.org opens the paper with a table of contents, a reading-progress indicator, and a chat panel. The same pattern works for bioRxiv papers on talk2biorxiv.org.
Is talk2arxiv free to use? What license does it have?
talk2arxiv is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is talk2arxiv still maintained?
Updated within 6 months. The last commit to talk2arxiv was on Oct 2, 2026.
What can I use talk2arxiv for?
Getting a quick summary of a new arXiv paper before reading it closely. Asking clarifying questions about a method, equation, or limitation in a paper. Checking a paper's claims against its evidence while reading. Exploring the work of a paper's authors and the papers it cites.
How do I install talk2arxiv?
Install dependencies: yarn Copy the example env file and add your OpenRouter key: cp .dev.vars.example .dev.vars Start the dev server: yarn dev
Where can I get talk2arxiv?
The source code and installation instructions are on GitHub: https://github.com/evanhu1/talk2arxiv

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects