7 devlogs
3h 57m 35s
A minimalistic yet feature-rich OCR (extract text from an image) software that just works - no ads, no login, no stored data and no bullshit.
Shipped this project!
ITS ON!!!
Please test it (it has some cool features for us, developers)
The worst part was definetly setting up Modal, but besides that it was fun
Thx!!!
completely changed the UI. i think it looks cooler now
0
Log in to leave a comment
It is now usable. The API connects with the front-end and extracts the text from the img. Response modes are also available
0
Log in to leave a comment
The API is 100% functional now. Now just need to connect it to the client, fix the 1000000 warnings I received and we can finally start tweaking the front-end and adding the small features
0
Log in to leave a comment
Finally managed to make flash-attn install. Used a pre-built version instead of compiling and building it locally (it would take hours). Now just have to fix some issues w/ huggingface transformers and we should be good to go
0
Log in to leave a comment
I’ve never seen a DX THIS BAD. Every library or version change I make means +5 minutes of build time. I’ma just let Claude handle the errors and go take a shower at this point
0
Log in to leave a comment
Right now, trying to deploy Deepseek-OCR (https://huggingface.co/deepseek-ai/DeepSeek-OCR) on Modal. It’s not my first time using Modal, but it has been a long time since I last used it, so I don’t remember the library’s syntax
0
Log in to leave a comment
Improved the UI on desktop. before, all elements were stacked exactly like on mobile. Now it looks better
0
Log in to leave a comment