A compliments app for mobile, built with React Native (Expo) and Django REST Framework, with a PostgreSQL database hosted on Railway.
Where the idea came from
The idea goes back to when I signed up at BeCode, in 2019. We had to suggest three personal projects, and I wrote down “compliment generator”. I never built it.
Back then, no language model was available to the general public. I would have had to write hundreds of compliments by hand, then hack together a system to match them to each person's traits. A lot of work for a result that was bound to feel repetitive.
Six years later, when I had to choose my final-year project, ChatGPT was in everyone's hands. The lock had come off. With a well-built prompt, an AI could write a unique compliment for each person, based on their profile. It was time.
The name comes from Hervé. Every Hervé is kind — and so is my compliment generator.
Each user builds a rich profile (physical traits, passions, background, life goal…) and gets a unique compliment every day, written by an AI (Groq API, LLaMA model). Users can also send compliments to each other, to a friend or to the whole community through a public feed.
Everything goes through an AI moderation pipeline. Each compliment is analysed when it is submitted, then approved or put on hold (insults, racism, harassment, spam). Borderline cases are reviewed by an admin in an Electron/Next.js desktop app. On the third warning, the account is suspended, and the whole app switches to a sarcastic, degraded mode.