Roasting HP is a project I made for fun. I was actually inspired by other earlier websites like Github Profile Roaster, and especially when I found PC Specification Roast. I thought it was a fun idea to roast phones too, so I made Roasting HP.
Development Journey
I actually didn't know how to make it at first, and then I found that they were using Gemini AI API that is free for some limit. So I decided to use it too. I didn't know the code, but someday (I don't remember when), I was trying Project IDX, which is an online code editor by Google, and I found the template project that was using Gemini API, and even better, they were using LangChain too. So I decided to use it as a base for Roasting HP. First, the code was just in vanilla HTML, but then I decided to migrate to React for better performance and maintainability.
Success & Challenges
When I finished making it, I tried to post it online and it was a hit. People loved it, and I even got a lot of views on TikTok. But, the limit was reached, and people got mad because it didn't work. I tried to figure out an easy solution and I just changed to another account API whenever someone reported it didn't work, but then I realized it wasn't effective. And then when I found Groq API, they provided many models that I could use for free, and the limit was very generous. Also, their streaming speed was faster than Gemini. So I added some models from it. But it's not really good in Bahasa Indonesia, especially when roasting using slang language - it feels too robotic. But that's okay, it's just for fallback when the Gemini hits the limit. Also, I added 2 Gemini models with separated limits which are Gemini Flash and Pro.