I vibe-coded an app that turns any recipe link into a nutrition label with Gemini + VS Code

I recently spent a month vibe-coding a web app that turns any recipe into a detailed nutrition label. My entire workflow was a loop of prompting Gemini, pasting the code into VS Code, fixing what broke, and repeating that x1000 times. I know about tools like Cursor & Replit, but I stuck with this manual setup since I'm more comfortable with it. No idea if it’s the “right” way, but I just ran with it and got this app launched.

You can give it a link to a YouTube video, recipe website, ingredients text, upload an recipe image (or take a photo of your recipe book), and it instantly:

• Analyze any recipe from a YouTube video, website link, pasted text or image.

• Identify ingredients and their quantities, then provide nutritional information obtained from USDA database.

• Generate a nutritional label, including calories, macro/micronutrients, and daily value of the recipe.

Feel free to try it here - https://recp.ai

My goal right now is to figure out where to take this next. Does this solve a real problem for you? What’s one thing you liked or disliked? What's the single most important feature I should build next?

Personal challenges so far:

• Getting transcripts reliably from YouTube is also a pain in the ass. I originally obtained them using a library with multiple proxies, then it didn't work for a while and I had to immediately find workarounds to fix it which took me a full day.

• Gemini API cost would mount up over time as I am using it for the heavy lifting (parsing ingredients, quantities, transcript summary into cooking steps, etc.). To mitigate this, I've implemented caching for recipe URLs, but the costs for unique, new analyses could still add up quickly.

Future goals:

• The next idea is to level-up the image feature. Instead of just OCR, I want it to recognize the actual dish from a photo - accurately detecting 'lasagna' or 'beef pho'. From there, it would identify the ingredients and run it through my database and USDA calculator. I've seen other apps attempt this, but honestly, their accuracy feels a bit off. I got three different calorie counts (400, 600, and 800) for the same photo on one of them, so I'm skeptical about the accuracy - this is what I am looking to solve.

• And if this app actually takes off, my vision is to build it into a full-on personal nutrition dashboard. You could save recipes, log meals, and track it all against your goals (like TDEE), with the app handling all the analytics, and providing you with daily reminders. That'd probably have to be a subscription based service (SaaS). Is that something you'd find useful enough to pay for?

Thanks so much for reading, feel free to ask me anything, any suggestions are welcome!