---
id: 36
type: blog
title: "From Thai Draft to English Technical Documentation"
slug: "thai-draft-to-english-technical-documentation"
canonical_url: "https://docs.phatysd.me/content/thai-draft-to-english-technical-documentation"
status: published
revision: 2
published_at: "2026-09-19T16:34:11.292Z"
updated_at: "2026-09-19T16:34:11.292Z"
hierarchy_path: "Web Blog > Engineering Journal > Documentation"
topics:
  - "Technical writing"
  - "Localization"
  - "AI agents"
  - "Documentation quality"
tags:
  - "writing"
  - "localization"
  - "agents"
  - "english"
  - "documentation"
related:
  - slug: "what-i-learned-writing-nextjs-guide-agents"
    title: "What I Learned from Writing a Next.js Guide for Agents"
    type: blog
    canonical_url: "https://docs.phatysd.me/content/what-i-learned-writing-nextjs-guide-agents"
  - slug: "building-agent-readable-knowledge-site"
    title: "Building an Agent-Readable Knowledge Site"
    type: blog
    canonical_url: "https://docs.phatysd.me/content/building-agent-readable-knowledge-site"
---

# From Thai Draft to English Technical Documentation

The first version of the documentation was written in Thai because that was the fastest way to capture the operational context. The published audience, however, includes coding agents and developers who expect English technical vocabulary, predictable headings, and searchable source references. Translation was therefore an information-architecture task, not a word-for-word replacement.

## Preserve meaning before style

The first pass kept the technical nouns stable: Server Component, Client Component, Server Action, Route Handler, `proxy.ts`, `DATABASE_URL`, `AUTH_URL`, revision, canonical URL, and cache tag. Translating or paraphrasing these identifiers would make code search and troubleshooting worse.

The prose around them became English, while code, commands, environment names, and URL paths remained exact.

## Make headings searchable

Headings were rewritten as direct English concepts:

- “Agent summary” instead of a general introduction.
- “Public content contract” instead of a broad explanation of the site.
- “Debugging a localhost redirect” instead of a narrative-only bug title.
- “Verification checklist” instead of an implied conclusion.

This makes table-of-contents navigation and retrieval more useful to both people and agents.

## Keep examples executable

Every code block should be valid enough to communicate the intended API. Comments should explain the boundary in English, not hide an important rule in a language the target agent may not parse consistently. Placeholder credentials remain placeholders; production secrets never belong in examples.

## Use source maps, not authority by tone

English prose can sound authoritative even when it is only a recommendation. Each framework-sensitive section links to official Next.js, Auth.js, GitLab, Cloudflare, Heroku, or Drizzle documentation. Repository-specific behavior links back to the actual contract or source file.

## Visual QA still matters

Text correctness does not guarantee readable rendering. Fenced code blocks need a code-specific CSS rule because an inline-code background can override the block style through selector specificity. After translating the article, the page, Markdown endpoint, JSON endpoint, and deployed CSS were checked again.

## Takeaway

Good technical localization preserves identifiers, improves retrieval structure, keeps examples executable, and verifies the rendered result. The output is not merely translated text; it is a more reliable interface for the next reader or agent.

## References

- [Next.js documentation](https://nextjs.org/docs/app)
- [Auth.js documentation](https://authjs.dev/)
- [GitLab OAuth documentation](https://docs.gitlab.com/integration/oauth_provider/)
- [Cloudflare Error 525](https://developers.cloudflare.com/support/troubleshooting/http-status-codes/cloudflare-5xx-errors/error-525/)
