Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Show HN: Extend UI – open-source UI kit for modern document apps (extend.ai)
233 points by kbyatnal 20 hours ago | hide | past | favorite | 68 comments
We're open-sourcing 14 components & examples today for PDF, DOCX, and XLSX viewers, plus bounding box citations, file upload, e-signature, and more. It's MIT licensed and fully customizable.

Demo video here: https://share.extend.ai/kRmSGKRF

When we started, we tried every file viewer and document component library we could find. Unfortunately, none of them had all the functionality (and polish) that we wanted, so we ended up building our own for https://extend.ai/. It was only ever meant to be internal, but enough customers kept asking for it that we decided to open source it.

It's useful for building document processing agents, real-time user facing document intake flows, or all kinds of internal tooling.

We naively thought this would be a solved problem. Turns out, making PDF/XLSX/DOCX viewers that work at scale is not trivial...we use and maintain it for Extend ourselves, so we've fixed a lot of edge cases that came up while running millions of pages / day through our own system. Our hope is that with our resources + community support, it'll keep getting better over time.

 help



The performance is quite bad. There are some basics missing or broken.

- no caching for file previews

- no sorting by date in the file picker

- no sorting by size in the file picker

- no sorting by file type in the file picker

- no search in the file picker

- cant enter a folder in the detail view (only expand) in the file picker

- cant go to page (by page number) in the document viewer

- after clicking a button in the document viewer, focus is lost on the document and arrow keys, space, pgup/pgdown dont work until the document is clicked again

- cant select text in the document viewer, unless I search first, in which case it then works

This is after looking at the file picker and document viewer less than 3 minutes(!!!). I gave up after that. Getting 80% of a file picker is easy, getting the last 20% done, so that it's on par with existing software, is not.

Very very odd to have things look this polished, yet be this terrible functionality- and performance-wise. These are not random quirky new ideas I'm having, these have been basics that work in every single file picker and document viewer since the early days of UI, before the web.


Your issues are mostly with the file picker, which is a brand new component that was just added yesterday in response to a feature request after launch: https://x.com/andrewlu0/status/2064726998186971400?s=46

It has some rough edges as a result, but your suggestions are very valid. We’ll make those improvements (we’re also taking PRs!).


Yeah, it also seems it's made in React & Next. From experience, you'll have a hard time choosing a worse technology for low-level, interaction-heavy UI. You need direct control over focus, keyboard handling, scrolling and so on. You also need to leverage hardware accelerated rendering too, which is clunky with 3rd party React libraries.

What's more, even if state management should technically be easier with the amount of state libraries, you'll realise sooner or later that the established ones are cleverly immutable where you really just want them to be performant.

I am not saying that it's React at fault for the symptoms you see here, but I would expect any such library made in it to hit exactly these kind of edge cases.


> From experience, you'll have a hard time choosing a worse technology for low-level, interaction-heavy UI.

Can you elaborate on that? What else would you recommend?


Yes! First, you need to get the DOM-level behaviour right (dragging, resizing, cursor focus). That could mean you may use or build primitives like with interact.js, CodeMirror/Monaco (realistically just use, not build), Fancytree. I really like Muuri for layouts. Usually there's also subtle interaction physics going on when building such libs in unexpected places (such as resizing and drag N drop speed curves), which you should tweak to your taste.

For framework, React wants to own and mediate the DOM via virtual tree, which is a major bottleneck when you need direct control over focus/selection/keyboard routing or hardware accelerated canvases. Instead, look at Svelte or Solid.js, as they integrate nicely with imperative DOM-oriented JS libraries and don't require heavy wrappers or indirect references for the 'unfriendlier' DOM nodes like canvases, scroll containers and so on.

If you're building an OS-like UI, you should also care about state and be sure you have direct control over where your data lives. For example, I usually build with Solid.js and a mix of custom object and lifetime code plus Solid stores for reactive surface state.

I usually end up managing object lifetimes because I end up needing to handle messy edge cases around reference vs value semantics and state merging (e.g. keep cursor position sane after a file sync or track focus across multiple windows, especially after refresh)

For text editing, if you use Monaco, it has so many internal lifecycle hooks you want to be aware of and interact with directly, that you'll see that most of implementation will end up outside classic frontend lib fast and I'd rather build the thing instead of bridges and wrappers to talk with a high-level framework.

All in all, you probably want to own a lot of state and behaviour yourself, and add a cooperating framework on top instead of an all-encompassing one.


Hey! Very valid feedback - working on improving the file system component and overall site performance today

Since there was a recent thread on react compiler[r] - I wonder if adding/pushing the code through react compiler would help? (Assuming it's not already being used)

[r] Thread was about rewrite in rust, but it made me have a look at the purpose/claims made by the project - and fine-grained, automated memoization for speed seems central.


its nice, people are harsh, you did something useful, thank you.

Looks cool but your home page heavily lags on my mbp m3 pro - you should maybe be lazy loading vs loading all your components upfront

Valid feedback a few other comments have mentioned , I’m going to try improving the home page tomorrow

The performance of the "Layout blocks" component is particularly bad. It consumes tons of CPU when scrolling.

Super cool. Working on a local AI tool specifically for document workflow automation (where context = screen/web/folders/files), and this could come in super useful. I do most of the PDF/DOCX/etc. parsing natively in Rust, but having a nice way to see the output without spinning up Word or Powerpoint is a huge leap.

Thanks for releasing publicly.


nice - did you write a custom parser for PDF/DOCX? we wrote one for XLSX after running into event loop issues with sheet JS

Using lopdf[1] for PDF parsing, rtf-parser[2] for RTF, calamine[3] for XLSX, and I'm sure you know that DOCX/PPTX/etc. is basically just a zip file of XML + text. The LLM cares about textual data (which just gets moderately cleaned up post-extraction), so I (thankfully) didn't have to deal with rendering. But showing a preview or end-result to a user would be a huge plus, so I can see myself using your library.

[1] https://github.com/J-F-Liu/lopdf

[2] https://github.com/d0rianb/rtf-parser

[3] https://github.com/tafia/calamine


What about rendering? That's the hard part.

we built a library @extend-ai/react-xlsx on top of it that renders the parsed contents onto a canvas

testing was mostly manual with a test corpus we generated. its not perfect but its pretty close for most files we've seen


For me, rendering was just a nice-to-have.

Sorry I meant to ask the author of Extend UI not you.

First of all thanks for the great library. It is so much more thay an UI kit!

We wrote (should say are writing) our own xlsx parser in Rust on IronCalc:

https://github.com/ironcalc/IronCalc/tree/main/xlsx


Experienced the same nightmare a few times over with citations.

React/next is limiting, we have rebuilt this for angular and now redid it again using lit for better compatibility. Our old one is very similar to this.

Loading citations for each field across 1000s of pages, colliding citations for all the messy formats, zoom, rotate etc. what a mess!

Great that you took the time to MIT this as it would have saved us many hours, though I think today Fable + Codex makes it pretty quick

I could recreate these in lit as a fork, would be very useful to have the full set


Someone requested web components earlier so let me know if you make that happen!

The performance on the site is truly terrible, the top animation was probably going at 5fps when scrolling.

Heard the performance feedback will work on it today!

Those bounding box demos are decent.

By quirk of fate i've spent the past 2 days prototyping some stuff on pdfjs. Just trying to figure out a game plan for handling bounding boxes in the face of page zooming, different resolutions etc. etc. I can't see it mentioned whether the components are virtualising pages (as in reusing dom elements as document pages scroll by). I guess i just learned what i'll be exploring tomorrow then...


yes - the pdf/docx viewer use react-virtual to virtualize the pages

the zoom should work with the bounding box highlights, we're working on adding rotation support


Why doesn't it mention anywhere that they are React components?


Hey! Those components are used for the demo site, and shouldn’t ship or get installed when you install the components in the registry.

None of the ones in the registry should have next as a dependency, please lmk if otherwise


React (vs web components) might be questionable, but Next.js as a hard dependency, for a component library, is just weird.

Responded above but the components like the file viewers should not depend on next. Please lmk if it gets pulled when installing via CLI

we'll try to make it more clear on the landing/introduction page!

_Try_ to? Are you serious? We're not taking ambiguity due to phrasing, we're talking straight up not mentioning it on either the front page, on the show all page or even on the page the components button takes you to. Not even on this forum thread. There's barely any text on the front page to change and there are even multiple places you can mention React. You could mention it in the title, e.g. "Open source _React_ UI kit for modern document apps" or you could add it to the second paragraph, e.g. "React components ready to drop into user-facing flows, agents, or internal tools." Not to mention the components page. And given what you use to make the website, how does it take this long to update something this trivial when you are going out of your way to promote it, replying to comments and your other services by proxy?

Also, add either user interactive loading of components or lazy load the demos, the amount of demos murder performance on phones.


coming back to this now and added it in a few places

on the demos - everything below the fold is lazy loaded but i will see what we can do to improve the mobile perf


Tone.

It looks really great! But, from what I can see, it’ll need some work on optimisation. Good luck with this project, guys!

Thank you! Working on optimizing

Looks great.

I'm curious whether the primary users are AI-native document products or more traditional SaaS applications.

The document-app niche feels increasingly important with the rise of AI workflows.


Cool project! I was playing around with the Excel viewer - the docs claim "Search across sheets and cell ranges", but I can't seem to trigger search functionality and the browser search bar can't find contents on cells.

Is this a known issue?


Mind sharing a link to the source code? i couldn’t find it on the website.


Github icon in the header on the right.

These should really be web components. Leaving out every framework other than React is really bad for the web.

Then build it yourself. They do work for free, give it to the world, and your response is: do it differently the way I want.

let's all try to be nicer

thats fair and definitely something we can try supporting in the future. we started with React because of how familiar models are with shadcn and tailwind

we hope this can be useful for people building in React though!


Thanks, that looks awesome! We were looking to add DOCX and XLSX preview to our app, and were planning to do server-side conversion to PDF (which seems to be what most other apps resort to) due to the lack of good libraries to render it, and this is exactly what we were looking for! :)

thanks! would love to get your feedback

i can't promise its visually 1:1 with Word/Excel but its pretty close on the corpus we tested with


really like these - curious how the xlsx editor and viewer is built in that what kind of headless spreadsheet?

could not have been easy


Hi! I'm one of the engineers at Extend that worked on this - one of our other engineers created a Rust XLSX/XLS parser that we ported to WASM to our @extend-ai/react-xlsx package which handles the rendering/charts. It exposes some hooks so users can use their own components for the toolbar

How much was actual engineering and how much was telling an AI what to do?

Even if it was just prompting, not sure how that takes away from the final, fairly polished, product. How do you define "actual engineering"?

How is your PDF coverage? They are notoriously difficult things to render, with endless edge cases. Mozilla’s PSD.js is the status quo here, so what do you do better than them?

We use react-pdf for the viewer which I believe uses pdf JS under the hood

We aren’t not trying to reinvent that engine, rather just provide a building block for people to plug in their design system to its controls



if you mean our docx viewer/editor specifically its hard to say without manually testing the visual fidelity with Word on some complex docx files

you are welcome to try it with your own documents and see, but its just one example we wanted to show. for the blocks that use the react-docx library, you can always copy the code and use a different method to render the docx file/thumbnails


so slow like molasses

“React components”.. I stopped there.

Looks clean and works fine, but it needs optimization. Clicking "Type" in the "schema builder" example takes 1~2 seconds to open the popover in the landing page(macbook pro m4). I think its because there are lots of heavy components, but still it's too slow.

it should be much faster on the individual component's page

the root page is a bit slow with all the viewers, in practice you probably wont have that many in your app on one page


Does it/will it support Markdown files?

i felt like we couldn't build much on top of react-markdown, which i think is what most people are using

This is really interesting. Thanks for creating this.

Excellent that you offer Miller columns, one of the best tools for computing and information browsing, and management. The world should run on Miller columns.

TIL those are called Miller columns !

> This page could not load

On mobile Safari…


hm seems to look ok on mine - is this on the root page?

good

I was excited by the headline but then when I got to the page

1. I saw it was React

2. Nothing loading, just a page full of spinners

I truly wish React could be launched into the sun.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: