FeedDoc – Extract a page range from PDF/PPT/Doc and attach it to AI chats

I built FeedDoc after getting frustrated with uploading entire documents

to AI chats when I only needed a few pages.

It's a Chrome extension with a side panel where you upload a PDF,

PowerPoint, or Word file, pick a page range, and it generates a new

sliced file and attaches it directly into the chat box of whatever AI

platform you're on — ChatGPT, Claude, Grok, Perplexity or T3 Chat.

The interesting technical bit: attachment works by reconstructing the

file as a File object in the content script and using the DataTransfer

API to set it on the platform's file input element, then firing the

appropriate change and input events so the React app picks it up. No

clipboard involved. For files over 50MB it falls back to

chrome.storage.session to avoid hitting Chrome's IPC message size limit.

PDF slicing uses pdf-lib entirely in the browser. PPTX manipulation

goes directly into the ZIP structure using JSZip since PPTX is just a

ZIP archive. DOCX page slicing is best-effort based on explicit page

break markers in the XML — true page count is render-dependent so that

one comes with a disclaimer.

Everything runs locally. No backend, no server, no data collection.

Files never leave the device.

Happy to answer questions on the implementation.

3 views