I built a Next.js + AI app that turns any book into a character map

Why I built it:

I love complex stories, but I always forget who’s who when I come back to a book.
So I built Booklaxy, a web app that generates interactive character maps and spoiler-free summaries using AI.

What it does

• 🧩 AI-generated summaries per chapter
• 🌌 Visual character relationships (interactive graph)
• 📝 Personal notes and book wiki

👉 Try it here → booklaxy.com

How I built it

Tech stack:
Next.js 14, MongoDB Atlas, Gemini AI, Tailwind CSS, GCP cloud run

What I learned

• Gemini handles multilingual books well, but prompt updates can break old results => always test to avoid regressions! (more complex that the basic tests)
• AI integration with the App is always heavier than it looks: design, async logic and retries add unexpected complexity.

Looking for feedback

I’d love to get feedback from fellow builders:

• How would you improve the onboarding UX?
• For those who’ve built prompt-based AI APIs, any tricks to keep things stable and efficient? :)