AI Prompt Gallery for 1M+ User Product
Built a full-stack prompt gallery and admin CMS for Plus AI's Nano Banana image generation feature — from database design to AI-powered content pipeline. 200+ prompts live at launch.
About Plus AI
AI-powered presentation tool for Google Slides and PowerPoint with 1M+ installations
Plus AI is one of the most popular AI presentation tools out there — over a million installations across Google Slides and PowerPoint, 900+ reviews, consistently rated 4.6-4.8 stars. Founded by Daniel Li, the product lets teams generate, edit, and redesign presentations using AI. Their Nano Banana feature generates professional images directly inside slides. The team saw an opportunity to capture search traffic from people looking for Nano Banana prompt ideas — but building that content at scale was a different problem entirely.
The Challenge
People were searching for Nano Banana prompt ideas, and Plus AI wanted to capture that traffic — an SEO-driven gallery of prompts that would rank on Google and funnel visitors into the product. The problem was scale. Creating each prompt page meant generating images, writing descriptions, crafting SEO metadata, and uploading to a CDN — all by hand. At that pace, building a catalog of hundreds of prompts would take months. The team didn't have the bandwidth to do it manually. They needed a production-ready gallery that users could browse and search, plus an admin tool that would let the team create new prompt pages in minutes instead of hours.
What I Built
I built the Plus AI Prompt Library — a full-stack application with a public gallery and an admin CMS. The public side is a browsable, SEO-optimized catalog of AI image generation prompts organized into 11 categories. Each prompt page shows the generated images, the exact prompt used, and a one-click button to use it in Plus AI. Behind it, a complete admin dashboard where the content team can generate images via fal.ai, auto-create SEO descriptions with Google Gemini, upload to bunny.net CDN, and publish — all without touching code. The whole thing runs on Next.js 16, PostgreSQL via Supabase, and Prisma ORM, with timing-safe HMAC authentication protecting admin routes.
Public Prompt Gallery
Browsable catalog with 11 categories, image previews, copy-to-clipboard prompts, and direct 'Use in Plus AI' CTAs.
AI Image Generation
Integrated fal.ai pipeline. Generate 1-4 images per prompt with 5 aspect ratios and 2K/4K resolution support.
SEO Auto-Generation
Google Gemini generates SEO titles, descriptions, and page content. Every prompt page is search-optimized automatically.
Admin CMS Dashboard
Full content management with category and image CRUD, stats overview, recent activity feed, and quick actions.
CDN Pipeline
Auto-upload generated images to bunny.net CDN with regional endpoints for global delivery.
HMAC Authentication
Timing-safe cryptographic authentication with signed sessions and middleware-protected admin routes.
ISR with On-Demand Revalidation
Static generation for performance and SEO, with instant revalidation when new content is published.
REST API Layer
Full CRUD APIs for categories and images with pagination, filtering, sorting, and CORS support.
Test Suite
Unit tests with Vitest, API integration tests, and E2E tests with Playwright covering critical user flows.
What This Unlocks
Time and cost spent on repetitive work, before and after.
| Task | Manual | With Plus AI | |
|---|---|---|---|
| Create a new prompt page | 30-45 min | 3-5 min | |
| Generate & upload images | 15-20 min | 1-2 min | |
| Write SEO metadata | 10-15 min | 0 min (auto) | |
| Add a new category | 1-2 hours | 2-3 min | |
| Publish & go live | Manual deploy | Instant (ISR) |
Create a new prompt page
Generate & upload images
Write SEO metadata
Add a new category
Publish & go live
What used to take 1+ hours per prompt now takes under 5 minutes. At 200+ prompts, that's 200+ hours saved at launch alone.
The Solution
I built the whole thing end to end in 3 weeks. A public-facing gallery with 11 categories, browsable prompt pages, and full SEO — so every prompt ranks on Google. Behind it, an admin CMS where the team generates images with fal.ai, auto-creates SEO descriptions with Gemini, and uploads to bunny.net CDN — all from one dashboard. The database runs on Supabase with Prisma. Admin routes are locked down with timing-safe HMAC authentication. The public side uses ISR with on-demand revalidation, so new prompts go live instantly without redeploying. I wrote the API layer, the image generation pipeline, the SEO automation, and a test suite covering unit, integration, and E2E. Shipped with 200+ prompts across 11 categories on day one.
Delivered as a production-ready Next.js application deployed on Vercel, with a Supabase database, Prisma ORM, and integrations with fal.ai, Google Gemini, and bunny.net CDN. Includes full admin interface, API documentation, and a comprehensive test suite. The client's content team was creating new prompts independently within the first day.
- Production-deployed on Vercel with custom subdomain
- Admin CMS — no developer needed to manage content
- AI-powered image generation and SEO pipeline
- 200+ prompts seeded across 11 categories at launch
- Comprehensive test suite (unit, integration, E2E)
- Content team onboarded and self-sufficient on day one
Results
Want results like these?
Whether it's a workshop for your team or a system built from scratch — let's talk about what AI can do for your business.
Get in touch