Research and LaTeX review: PDF region captures with comments in one prompt

[oedd]

Researchers and students often work with an agent on a paper, thesis, or compiled LaTeX draft. It would be useful to point at an equation, figure, or paragraph in the PDF, add a comment, and collect several such notes into one prompt. Today that involves manually taking screenshots and keeping each image paired with its comment and page number.

The smallest useful feature would be:

• A dedicated PDF preview in the right sidebar beside the conversation.
• Drag a rectangle to capture part of a page; a nearby annotation popup offers a comment, Attach, and delete, matching the web annotation interaction.
• Repeat across pages and send the screenshots and comments together through the existing composer.

This does not need semantic block selection, OCR, PDF editing, or annotation export. The source PDF can stay untouched. The goal is simply to make reviewing a compiled document with an agent as convenient as annotating a web preview.

A prototype is available in my fork, based on v0.0.42. It uses PDF.js for rendering and reuses T3's existing region-annotation and image-attachment flow. It renders one page at a time, loads the reader on demand, and bundles its fonts/decoders locally. It works from the file panel and PDF attachment previews in web/desktop; mobile and PDFs opened explicitly in the integrated browser are outside this prototype.

Short demo

The current reader has page navigation and zoom, but no text search, text selection, or continuous scrolling. Each screenshot uses one of the existing eight attachment slots. These are deliberate prototype limits, not a proposal for a full PDF editor.

Would this research/LaTeX workflow fit the project's direction? I understand the contribution guidelines and am happy to maintain it in my fork or adjust the scope based on feedback. Draft PR #13874 is a concrete implementation reference, without an expectation that maintainers merge it.

Implementation assistance: GPT-6 via Codex.