RecapAi

Inspiration

Meetings produce information — they rarely produce action. After almost every call, decisions get made and tasks get assigned verbally, but within a day half of it is forgotten because nobody wrote it down properly. We wanted to build something that removes that friction entirely: turn a raw conversation into a structured, shareable record of what was decided and who owns it — automatically.

What it does

RecapAI takes a meeting transcript and, using generative AI, produces:

• A clear, concise summary of what was discussed
• A structured task list, with each action item automatically assigned to the right person

Users can create an account, paste a transcript, generate a summary in seconds, and check off tasks as they get completed. Free users get 3 meetings/month; a Stripe-powered Pro subscription unlocks unlimited use.

How we built it

• Frontend/Backend: Next.js (App Router) with TypeScript and Tailwind CSS
• Database: PostgreSQL (Supabase), modeled with Prisma across User, Meeting, Summary, and Task
• Authentication: Auth.js with hashed credentials and JWT sessions
• AI: Groq API, with a provider-agnostic architecture designed to swap models (Groq, Gemini, Claude) without touching business logic
• Payments: Stripe Checkout + webhooks to upgrade a user's plan automatically after payment
• Deployment: Vercel, with environment-based configuration for local vs. production

We also designed a custom visual identity (typography, color system, a dedicated landing page) so the product feels like a real SaaS rather than a prototype.

Challenges we ran into

• Groq deprecated the LLM model we originally integrated mid-build, requiring us to adapt the AI layer on short notice
• Getting Stripe webhooks working locally (via Stripe CLI) and then correctly reconfigured for production on Vercel took careful attention to environment variables
• Designing prompts that reliably return clean, parseable JSON from the LLM — rather than inconsistent free text — took several iterations

Accomplishments that we're proud of

Shipping a complete, working SaaS loop end-to-end: sign-up → AI-generated summary → payment → plan upgrade — all live in production, not just a local demo.

What we learned

How to architect an AI feature around a swappable provider layer, how Stripe's webhook-driven billing model works in practice, and how much of a difference a consistent visual identity makes to how "real" a product feels.

What's next for RecapAI

• Google Meet / Google Drive integration to auto-import transcripts
• Audio upload with automatic transcription
• PDF export of summaries
• Automated tests and a CI pipeline

Built With

• auth.js
• css
• groq
• next.js
• postgresql
• prisma
• stripe
• supabase
• tailwind
• typescript