Launching today

Astra Annotator

Launching today

Turn PDFs into labeled text, figures and equations with AI

1 follower

Turn PDFs into labeled text, figures and equations with AI

1 follower

Visit website

Data visualization tools

•

Automation tools

Tell AI what to find in your documents, then watch editable annotations appear across up to three pages at once. Separate text, figures, tables and equations; add your own labels and notes; export the parts you need. Works with PDFs, Office files and images. Try the paper demo without a key, bring your own API key in the browser, or run locally with Codex App Server. Desktop-first. English, Japanese and Chinese UI. AI usage is billed by your provider.

Launch tags:Productivity•Developer Tools•Artificial Intelligence

Launch Team / Built With

OpenAI

GPT-6 Astra Challenge

What became possible in your product with Astra that was not practical before?

Maker

Astra made a unified, instruction-driven annotation workflow practical for this app. A user can describe what matters in a document and the agent returns editable regions with text, labels and notes, rather than requiring a separate preset for every task. The recorded demo uses GPT-6 Astra through Codex App Server to analyze three pages of an OpenAI paper concurrently, separating text, figures, captions, tables and equations. Annotations arrive in the workspace as the run progresses; users can correct them, add their own regions, enforce exact label names, and export selected content. This combines visual interpretation, structured extraction and human review in one document canvas. The GitHub Pages version also supports user-configured API connections, while the saved paper demo can be explored without an API key.

Report

6h ago

Maker

📌

Hi Product Hunt! 👋

Astra Annotator turns a plain-language instruction into visible, editable document annotations. Describe what matters, watch AI label regions across up to three pages at once, then refine the result by hand.

Try it in under a minute:

1. Open the app on a desktop browser.

2. Click ‘Explore a paper’ to inspect saved annotations on an OpenAI research paper. No API key is needed for this demo.

3. Select a region to inspect its text, label or equation, or draw your own annotation.

For a fresh AI run, upload a document and configure your API connection, or run the local Codex App Server version. Try this instruction:

‘Separate the title, body text, figures, tables, equations and captions. Preserve the original wording, return equations as LaTeX, and flag unclear content for review instead of guessing.’

You can let AI create labels or supply exact names and definitions. Export annotated PDFs, image crops, text and structured data. The GIF shows a real GPT-6 Astra run through local Codex App Server; opening the saved demo does not trigger AI inference.

A few practical details: the app is free to use; AI usage is billed by your provider. In browser-only mode, documents are processed in your browser and page images are sent to your configured API when you run AI. Your API key stays in the tab’s memory. Browser Office previews focus on extracted text; use the local document server when original Office layout matters. Review AI output before relying on it.

Try it: https://ryusui-hiro.github.io/DocumentAnnotationAgent/

Source, setup and issue reports: https://github.com/ryusui-hiro/DocumentAnnotationAgent

What would you annotate first: a research paper, a contract, or a spreadsheet? Which part takes the most manual work today?

Report

4d ago

Forum Threads

p/astra-annotator

•

4d ago

What should AI extract from procedures, work instructions, manuals, and papers?

I m building Astra Annotator, initially for procedures (SOPs), work instructions, manuals, and research papers. I plan to broaden the focus to Office document workflows later.

Early-release note: The app has only recently become available. Bugs and rough edges remain, and AI may miss, misread or mislabel content. Please treat results as draft annotations, check them against the original document, and report problems through GitHub Issues. The Product Hunt launch is scheduled for September 18.

The idea is simple: describe what matters, watch AI place editable labels and notes on the page, then extract the useful parts. You can also mark regions yourself or supply exact label names.

Possible ways to use it

View all