Back to browse
Prompt-to-Excalidraw demo with Gemma 4 E2B in the browser (3.1GB)

Prompt-to-Excalidraw demo with Gemma 4 E2B in the browser (3.1GB)

by teamchong·Apr 19, 2026·163 points·62 comments

AI Analysis

●●●BangerWizardryBig Brain

50-token compact code output beats raw 5,000-token Excalidraw JSON — clever compression.

Strengths
  • WGSL compute shaders hit 30+ tok/s GPU inference entirely client-side.
  • TurboQuant polar + QJL algorithm compresses KV cache 2.4× for longer conversations.
  • Sibling npm package offers same algorithm in WASM+SIMD for CPU vector search.
Weaknesses
  • Desktop Chrome 134+ only — no Safari, iOS, or broader browser support yet.
  • Requires ~3GB RAM, excluding most mobile browsers from running the demo.
Category
Target Audience

Frontend developers, ML engineers interested in edge inference

Similar To

WebLLM · Transformers.js · MLC Chat

Similar Projects

AI/ML●●Solid

Agent skill for Excalidraw diagrams with animation and image

Auto-animated Excalidraw diagrams from prompts beats manual canvas setup.

Ship ItSolve My Problem
muthuishere
102mo ago