Mastodawn

Steve Krouse (@stevekrouse)

Cerebras의 GLM 4.7 모델을 이용해 초당 약 2k 처리량으로 HTML을 최대한 빠르게 생성하는 실험을 소개합니다. 부모 에이전트가 슬롯(slot)을 포함한 HTML 스캐폴드를 생성하고, 각 자식 에이전트가 할당된 슬롯을 병렬로 채워 최종 페이지(예: initializeScene 함수 등)를 분할·병렬 생성하는 멀티에이전트 기반 파이프라인 방식입니다.

https://x.com/stevekrouse/status/2020970356413005865

#cerebras #glm #htmlgeneration #multiagent #parallel

Steve Krouse (@stevekrouse) on X

experimenting generating html as fast as possible 1. using cerebras glm 4.7 @ 2k tps 2. parent agent generates html scaffold with "slots" 3. each child agent is assigned to generate slot /* SLOT_START:initializeScene */ function initializeScene() { return ""; }

X (formerly Twitter)

Angel Munoz Jan 23, 2024

After quite a while, I'm fairly sure that Hox will handle most of your HTML trees 😆

I was able to produce a 1.14gb html file with async code in-between maybe not the fastest code ever BUT for usual cases it should be good enough :)

I will make a release soon!

Here's the repo :)

https://github.com/AngelMunoz/Hox

#dotnet #fsharp #htmldsl #htmlgeneration #backend #webdev

GitHub - AngelMunoz/Hox: Async HTML rendering and a simplistic but extensible DSL based on css selectors.

Async HTML rendering and a simplistic but extensible DSL based on css selectors. - GitHub - AngelMunoz/Hox: Async HTML rendering and a simplistic but extensible DSL based on css selectors.

GitHub