LazyChef & Pantry Autopilot 🦥

Description and use

LazyChef is a local‑first kitchen copilot that uses Gemma 2B and WebGPU to instantly match and generate recipes from pantry items, enforcing one‑pan, ≤3 steps, ≤15 min constraints, while tracking lazy payoff metrics.

Problem

Developers and neurodivergent users face cooking decision paralysis and dirty pan dread after long coding sessions, leading to expensive takeout and wasted ingredients.

Outcome

LazyChef delivers a zero‑latency, privacy‑preserving cooking assistant that reduces decision fatigue, saves time, and cuts food waste, all within a 350 KB gzipped bundle.

Workflow

• Select pantry items in the UI
• Choose laziness level (One‑Pan Only or Speed Demon)
• Instant heuristic engine matches recipes in ≤15 ms
• On‑device Gemma 2B generates custom variations in a Web Worker via WebGPU
• Display step‑by‑step cooking assistant
• Track lazy payoff metrics (minutes saved, pans saved, money saved)
• Persist pantry and stats in IndexedDB for offline use

Evidence

Instant heuristic engine matches recipes in under 15 milliseconds

Source

On‑device Gemma 2B open‑weight model runs locally to synthesize custom creative variations

Source

The entire production bundle is kept under 350 KB gzipped

Source

Zero telemetry. Not a single byte of your pantry data or cooking habits leaves your device.

Source