An AI-driven restaurant simulation where LLM-controlled rats run a kitchen with personalities, needs, and autonomous decisions.
Early alpha: this is currently a single-day replay build. Bugs and weird behavior are expected. Multi-day learning + bring-your-own-API runs are coming next.
built by putt · inspired by smallville
This is best on laptop/desktop. Mobile works, but spacing and controls are tighter.
What's been built and what's coming next
3 AI rats (Basil, Sage, Noodle) with personalities, needs, and autonomous LLM decisions running a full restaurant service loop.
A* pathfinding, collision resolution, door claims, station contention, rotating priority, speed bonuses.
State machine customers (arrive → order → eat → pay → leave), patience/satisfaction, archetypes, tipping.
Multi-step cooking (fetch → prep → cook → plate), sauce pipeline, prepped ingredient substitution, 3 base recipes.
End-of-day LLM-generated diary entries, working memory with lessons, multi-day personality persistence.
Static replay viewer with event timeline, room navigation, night overlays, and music.
One AI rat starts alone. Hire worker rats, buy upgrades, expand the dining room, and build a 5-star restaurant from scratch.
10 purchasable upgrades: second stove, industrial dishwasher, walk-in pantry, extra tables, ambiance boost, hire slots. Upgrades install overnight and change the grid.
Hired workers run simple TypeScript policies — no LLM needed. Dishwasher washes, cook cooks, server serves. Basil writes custom SOPs in natural language to modify their behavior.
Workers level up (Lv1-5) with practice, gain speed bonuses. Morale affects performance — overworked rats slow down or call in sick. Basil can train and give breaks.
Rolling 7-day satisfaction score → 1-5 stars. Higher stars = more customers + premium menu unlock (Truffle Risotto, Lobster Bisque).
High-impact day events: inspections, breaker failures, supplier no-shows, and influencer rushes.
Friendships and grudges between workers affect coordination, burnout risk, and reliability.
Nightly debriefs update SOPs, staffing priorities, and upgrade goals across days.
Run multiple locations with shared cash, staff transfers, and shared inventory pressure.
Rival AI restaurants compete for demand, reputation, and staff in the same city.
Automated seasons with rankings and replay highlights to compare model/provider setups.