@tryhamster/gerbil 1.2.2 → 1.3.1

package/README.md

@@ -5,7 +5,7 @@
 <h1 align="center">Gerbil</h1>
 
 <p align="center">
-  <strong>A from-scratch WebGPU/WGSL inference engine. Text, vision, embeddings, speech — all native, on-device, in the browser and Node.</strong>
+  <strong>On-device LLM inference for the browser and Node. Text, vision, embeddings, and speech — all on the GPU via WebGPU/WGSL, private and offline.</strong>
 </p>
 
 <p align="center">
@@ -48,8 +48,7 @@ const { text } = await engine.generate("Explain recursion in one sentence");
 
 ## Why Gerbil?
 
-- **One native engine** — a from-scratch WebGPU/WGSL engine, pure compute shaders, nothing
-  extra to ship.
+- **One native engine** — pure WebGPU/WGSL compute shaders, nothing extra to ship.
 - **~90 KB gzipped** — the _entire_ multimodal engine. No heavyweight ML runtime; model
   weights stream from the HuggingFace Hub at run time.
 - **Multimodal, all native** — text, vision (image→text), embeddings, and speech run on the
@@ -79,9 +78,10 @@ After global install, use `gerbil` directly instead of `npx @tryhamster/gerbil`.
 
 ## Native WebGPU Engine
 
-Gerbil's product is a from-scratch WebGPU inference engine — pure WGSL compute shaders.
-It loads safetensors directly from the HuggingFace Hub (selective tensor download — skip
-vision towers you don't need) and runs the same code in the browser and in Node (via Dawn).
+Gerbil runs LLMs directly on the GPU — pure WebGPU/WGSL compute, no Python, no native
+runtime, nothing to install. The same code runs in a browser tab and in Node (via Dawn),
+streaming weights straight from the HuggingFace Hub and loading only the tensors a request
+needs — skip the vision tower you don't.
 
 ```typescript
 import { WebGPUEngine } from "@tryhamster/gerbil/gpu";