How I Built an AI Talking Bunny
Real-time conversation, Supabase on the backend, and the joy of making a character that talks back. Lessons from a fun side project.
Some projects start with a business case. This one started with a question: what if a bunny could talk back?
The idea
I wanted to build something conversational that felt alive - a character with a personality, not just a chat box. The Talking Bunny takes user input, runs it through an AI conversation layer, and responds in character.
Why Supabase
For the backend I reached for Supabase. It gave me auth, a Postgres database, and real-time subscriptions out of the box - so I could focus on the conversation logic instead of wiring infrastructure together.
Making it feel alive
The personality is everything. The system prompt defines the bunny's tone, and the response streams back to the UI with a typewriter effect, so it feels like the character is actually speaking.
- Choose the right defaults - auth and database from Supabase saved weeks.
- Stream responses for a natural, human feel.
- Nail the personality before scaling the feature set.
It's proof of what I enjoy most: taking a fun idea and making it real, end to end.