receipt app

after much pain, i have enabled zooming on images.

first, I tried to handle gestures myself, and I built functions tracking taps and zoom. i got zoom to barely work, and then when i realised zooming requires panning too, i tried implementing that. and then i got stuck.

i asked codex for guidance and was led to react-native-gesture-handler to handle gestures. that did not solve everything. i was able to pan zoomed images out to oblivion, and then had to figure out how to clamp translation so that the image always stays in bounds and blank space is not shown when not necessary. and then i had to figure out how to work the translation clamp with the zooming scale.

pretty big pain in the ass, but i did end up getting it to load onto ios. ive never developed anything for ios before and even though expo is cross platform the process is a lot more difficult compared to android. also if you dont have a mac its pretty difficult to build lol

had to also deal with some google oauth stuff for using ios that i didnt have to do on android

okay i updated the api and the sheets to set up categories based on regular accounting principles. also made sure that the outputs of the AI are fully normalized and set up to be the same format, because before it would output in different formats depending on how the receipt displayed it. also did more google auth hardening and vercel backend stuff

lots of API backend work - i focused on overhauling the backend auth to harden my api. before, my api was open to the public, so even if people didn’t have my openai api key, they could still spam my api to eat up my usage. but now I implemented google oauth requirement, and oauth only permits whitelisted accounts to access my app, so the api is fully gated behind my whitelist.

After some testing/feedback, i implemented multiselect allowing users to select more than one image at a time from gallery. and then also added buttons to let them preview them before sending them off.

I created a working prototype - this is basically the minimum viable product. it takes a photo of a receipt and uses the openai api to read it, and then it puts it onto google drive and a spreadsheet.