Rizzlr

Inspiration

Many people struggle with dating, not knowing what to say or how to keep conversations flowing. We created Rizzlr as a practice environment with infinite attempts where users can get as much practice as they want with dating conversations before talking to that special someone.

What it does

Rizzlr is an AI-powered dating conversation simulator with:

• 300+ unique AI personalities with realistic bios and images
• Tinder-style swiping with gender preference filters
• Graded conversations (S-F) with detailed feedback on engagement, flow, chemistry, and timing
• Conversation history to track progress and try different approaches
• Match management with reset and removal options

How we built it

• Frontend: Next.js 15, TypeScript, Tailwind CSS, Clerk auth, Framer Motion
• Backend: Supabase (PostgreSQL) with custom schema for conversations and grading
• AI: Groq API (Llama 3.3 70B) for responses, Stability AI for profile images
• Profile Generation: Python for calling imagegen and LLM apis, hybrid LLM/local approach for speed

Challenges we ran into

• Integrating Clerk auth with Supabase (UUID/TEXT type mismatches with Clerk's string IDs)
• Gender filtering bugs
• Profile generation and prompt engineering mistakes
• UI problems stacked on UI problems

Accomplishments that we're proud of

• Sub-second AI responses using Groq
• 300+ diverse, non-repetitive AI personalities
• Complete conversation system with grading and live feedback
• Infinite conversation resets, saving to user conversation history
• Comprehensive history to view past conversations with feedback
• Polished UX rivaling real dating apps

What we learned

• AI prompt engineering for consistency and diversity
• Chat database design patterns and indexing
• TypeScript's value in full-stack development
• Small UX details make big impacts
• Rizz is harder than it looks

Built With

• groq
• nextjs
• python
• typescript