出自幽谷

WeChat → HTML

2025-10-14

微信這種玩意最大的噁心就一個詞：封閉！微信公眾號就更是！
無 oEmbed 的嵌入 API，無法還原全文，禁止 iframe 打開，各種所謂圖片防盜鏈⋯⋯
從互聯網的本真視角，是個號，很微信，並不公眾。
之前用到微信公號文章轉論壇時，專門為自己機器寫過一個腳本，通過 getwc 直接抓圖片再轉論壇；但寫完後也沒用過。
今天學生發微信公號內容到論壇，問“論壇怎麼搬微信公眾號文章（）貼連結看著太不方便了”⋯⋯ 回覆了學生一段話：

大多數正常網站或鏈接都支持嵌入，但微信⋯⋯ 從來就不是正常玩意兒。
公眾號文章，直接複製粘貼，文字是可以的，但圖片會因為微信官方做的設置直接不能正常顯示。
也因為微信官方的設置，整個鏈接在論壇無法正常顯示。
這是個人對微信這玩意兒深惡痛絕的原因之一。

然後，現身說法，抱怨之外，要自己解決問題。

兩個小時的成果：通過 https://wx.bdfz.net/ 網頁，學生此後可以直接快速搬遷微信公號內容到論壇了。

授之以魚與漁，手冊：

1) Overview

WeChat → HTML is a Cloudflare Worker that turns a WeChat article URL (mp.weixin.qq.com/s/...) into a permanent, fast‑loading web page with all text and images preserved. It also provides one‑click copy for Discourse and a Markdown export.

Frontend: a minimal home page where you paste a WeChat URL. It shows a permanent link, a live preview, and copy buttons.
Backend: fetches the article, sanitizes HTML, rehosts images to Cloudflare R2, stores the final HTML in R2, and returns a canonical slug URL like https://wx.bdfz.net/<slug>.
Instant hit (“秒回”): repeated requests for the same URL return immediately via a short‑hash URL index.

2) Features

Full‑text extraction of #js_content
Image normalization & rehosting to R2
Clean, mobile‑friendly article page (dark‑mode aware)
Markdown export: GET /api/export/<slug>.md
One‑click copy: “full text + link” (HTML + plaintext) and “Markdown”
Instant return if a URL has been ingested before
Health check endpoint
Fallback image proxy /img?u=

3) Architecture

    text
    
    
  
Browser → Worker (/api/ingest) → Fetch WeChat page
                          ↓
                   Extract #js_content
                          ↓
        Upload images to R2 (bucket: blog-images)
                          ↓
     Render final HTML, write to R2 (bucket: wx-articles)
                          ↓
     Save URL→slug index for future instant hits
                          ↓
Return { ok, url: https://wx.bdfz.net/<slug> } + render preview

R2 keys

Article HTML: articles/<slug>.html
URL index: index/byurl/<sha1(url).slice(0,8)>.json
Images: wechat/<slug>/<date>_<sha1>.<ext>

4) Requirements

Cloudflare account + zone for your domain (e.g. bdfz.net)
Cloudflare Workers & R2 enabled
DNS record for subdomain (e.g. wx.bdfz.net) proxied through Cloudflare
Node.js on your machine (commands use npx wrangler)

5) Cloudflare Resources (bindings)

Create two R2 buckets and two environment variables:

R2 Buckets
- wx-articles → stores rendered article HTML & URL index
- blog-images → stores rehosted images
Environment variables
- PUBLIC_IMG_BASE — e.g. https://img.bdfz.net (public host for blog-images)
- SITE_TITLE — UI title, e.g. WeChat → HTML

These are bound inside the Worker as:

    text
    
    
  
env.WX_ART   -> R2 bucket wx-articles
env.WX_IMG   -> R2 bucket blog-images
env.PUBLIC_IMG_BASE -> "https://img.bdfz.net"
env.SITE_TITLE      -> "WeChat → HTML"

Routing: add a Worker route wx.bdfz.net/* for the zone bdfz.net.

6) Install & Deploy

    bash
    
unset CF_API_TOKEN CLOUDFLARE_API_TOKEN CLOUDFLARE_ACCOUNT_ID
npx wrangler login
wrangler whoami

Create (or confirm) buckets

bash

npx wrangler r2 bucket create wx-articles
npx wrangler r2 bucket create blog-images

Deploy

    bash
    
# From the project root
npx wrangler deploy

You should see bindings listed and the route wx.bdfz.net/* applied.

7) Using the Service

A) Web UI

Open:

text

https://wx.bdfz.net/

Paste a WeChat URL → click Convert
You’ll get:
- Permanent link: https://wx.bdfz.net/<slug>
- Live preview rendered from the saved page
- Copy full text + link (rich HTML + plaintext fallback)
- Copy Markdown
- Download .md

Auto‑submit shortcut

Open with a prefilled URL (auto‑submit once; query is removed to prevent flicker):

text

https://wx.bdfz.net/?url=https%3A%2F%2Fmp.weixin.qq.com%2Fs%2F...

If the same URL was processed before, you’ll see “Already existed — served instantly.”.

B) API Endpoints

`POST /api/ingest`

Body can be form or JSON:

    bash
    
    
  
# Form
curl -X POST -F 'url=https://mp.weixin.qq.com/s/...' https://wx.bdfz.net/api/ingest

# JSON
curl -X POST -H 'content-type: application/json' \
  -d '{"url":"https://mp.weixin.qq.com/s/..."}' \
  https://wx.bdfz.net/api/ingest

Success:

    json
    
    
  
{
  "ok": true,
  "url": "https://wx.bdfz.net/<slug>",
  "slug": "<slug>",
  "title": "Article Title",
  "existed": false
}

Already exists:

    json
    
{ "ok": true, "url": "...", "slug": "...", "title": "...", "existed": true }

`GET /<slug>`

Returns the rendered HTML page with sanitized content and rehosted images.

`GET /api/export/<slug>.md`

Downloads Markdown built from the saved HTML.

`GET /img?u=<http-url>`

Fallback image proxy if some uploads fail or exceed the per‑article cap.

`GET /health`

Basic health check: { ok: true, service: "wx-ingest" }.

8) Technical Notes

Extraction: finds inner HTML of <div id="js_content">...</div>.
Sanitization: strips <script>/<style>, event handlers (on*), and javascript: URLs.
Images:
- Reads data-src or src, normalizes to HTTPS.
- Uploads up to 40 images per article to R2 in parallel (concurrency 8).
- Excess images fall back to /img?u=... proxy.
Slug: slugify(title) + '-' + sha1(url).slice(0,8).
Index for instant hits: index/byurl/<sha1(url).slice(0,8)>.json.
Caching:
- Articles: Cache-Control: public, max-age=3600
- Proxy images: Cloudflare cache ~1 hour

9) Discourse Workflow

No Discourse config changes needed.

Use Copy full text + link and paste into the Discourse editor.
The clipboard contains HTML (rich) and plaintext (fallback).
Or just paste the permanent link https://wx.bdfz.net/<slug>.
Prefer Markdown? Use Copy Markdown or Download .md.

10) Operations

Re‑ingest / Update

Delete the article and index keys from wx-articles, then re‑run /api/ingest for the URL.

Purge Completely

Remove articles/<slug>.html
Remove index/byurl/<hash8>.json
(Optional) remove images under wechat/<slug>/ from blog-images

Logs & Tail