zeekay
712a10d0f3
The home now says who makes Zen LM, Zoo Labs Foundation, a 501(c)(3) non-profit, the two jobs it is chosen for, that Zen 6 and Zen 6 Flash are available now, and that Zen 7 is a research preview with a link to request access. The footer names the foundation, and "Qwen" leaves the site keywords. Posts marked "upstream: true" carry data-upstream on their page, card and read-more link, and scripts/check-zen.mjs fails the build when a name from zen.upstream shows anywhere else. The build stops measuring remote images: fumadocs fetched every one to read its size, and a single host answering 403 failed the build.
3.6 KiB
zen-blog -- AI Knowledge Base
Project: zen-blog Organization: zenlm Repository: https://github.com/zenlm/zen-blog Domain: https://blog.zenlm.org Last Updated: 2026-03-26
Overview
zen-blog is the Zen LM blog, built with fumadocs + Next.js (static export). Zen LM is the open model family of Zoo Labs Foundation, a 501(c)(3) non-profit.
Architecture
- Framework: Next.js 15 + fumadocs-mdx 11 + fumadocs-core 15
- Content: MDX and Markdown files in
content/ - Output: Static export (
output: "export") - Styling: Tailwind CSS 4 + shadcn/ui components
- Math: remark-math + rehype-katex (KaTeX rendering)
- Code highlighting: Disabled (rehypeCodeOptions: false) due to Shiki/math incompatibility
Content Format
Posts are stored as content/{slug}.mdx or content/{slug}.md:
---
title: "Post Title"
date: "YYYY-MM-DD"
description: "Post description"
author: "Author Name"
tags: ["tag1", "tag2"]
upstream: true # only on a post whose subject is another lab's release
---
44 posts came from the QwenLM blog this repository forked. They carry
upstream: true: their page gets <meta name="data-upstream"> and their
article and cards carry data-upstream. They name the releases they are about,
by the names those releases had. A find-and-replace in b3b6eb0 once rewrote
"Qwen2"/"Qwen2.5" to "zen" in them (Qwen2-57B-A14B read "zen7B-A14B"); every
such token was restored on 2026-09-30. Never rename another lab's model in them.
MDX posts can use custom components: <Figure />, <LinkButton />, <Video />, <Fullwidth>.
Markdown posts use standard markdown only (no JSX).
Posts with bare < in text or complex math use .md extension (not .mdx) to avoid JSX parsing issues.
Commands
npm run dev # Development server
npm run build # fumadocs-mdx → next build → scripts/check-zen.mjs out
npm start # Serve built site
Zen copy names Zen
scripts/check-zen.mjs reads every built page and fails the build on a name
from zen.upstream outside a data-upstream element. Upstream names belong in
two marked places only: an Architecture value (a loader string) and a License &
attribution section — plus the upstream-subject posts above.
Remote images are not probed at build time (remarkImageOptions: { external: false }): one image host answering 403 used to fail the whole build.
Deploy
blog.zenlm.org is served by the edge's staticFiles middleware from
s3://hanzo-sites/zen/blog (universe infra/aws/routes/sites.yaml, route
blog-zenlm-org), a copy of the Cloudflare Pages deployment it replaced. A build
reaches the site only when out/ is published into that prefix. The
gh-pages branch and .github/workflows/gh-pages.yml (a Hugo build) are
leftovers from before the migration and serve nothing.
Migration History
- 2026-03-26: Migrated from Hugo (PaperMod theme) to fumadocs + Next.js
- Converted 71 blog posts from Hugo format
- Hugo shortcodes (figure, button, video, example, fullwidth) converted to MDX components or markdown
- Chinese (.zh.md) translations not migrated (English-only for now)
- JSON case files moved to
public/cases/
Rules for AI Assistants
- ALWAYS update LLM.md with significant discoveries
- NEVER commit model weights (*.safetensors, *.bin, *.gguf, *.pt)
- NEVER commit symlinked files (CLAUDE.md, AGENTS.md, GEMINI.md, QWEN.md)
- NEVER create random summary files -- update THIS file only
Context
This file (LLM.md) is symlinked as CLAUDE.md, AGENTS.md, GEMINI.md, QWEN.md.
Part of the Zen AI family -- Clarity Through Intelligence