What do mid, idx and sn mean in a WeChat article link, and why are there two link formats?

WeChat article links come in two shapes. The long one, which you get from a browser address bar, an account's history page or most scrapers, looks like mp.weixin.qq.com/s?__biz=MzA5NTI3MjIwMA==&mid=2650412873&idx=2&sn=8c1f0e5a3d9b47e6a2c4f17b0d9e3a65&chksm=…&scene=21#wechat_redirect. The short one, which WeChat increasingly hands out when you share or copy a link, is mp.weixin.qq.com/s/ plus a short code. Both open the same article. The trouble starts when one data set mixes them: the same article counts twice, a join on URL misses, and a feed reader announces an old post again because its link changed form.

The four parameters that matter

Parameter	Example	Meaning
__biz	MzA5NTI3MjIwMA==	The Official Account; see what __biz is
mid	2650412873	One push (群发) from that account; unique within the account
idx	2	The article's position in that push: 1 is the lead article (头条), 2 the second, and so on
sn	32 hexadecimal characters	A signature that goes with the three above; WeChat checks it, so a link with a missing or altered sn usually fails to open

__biz + mid + idx pin down one article. sn makes the link valid. Everything else is session or tracking data you can drop: chksm (a checksum that changes from fetch to fetch), scene, sessionid, key, pass_ticket, #rd, #wechat_redirect. A clean long link keeps __biz, mid, idx and sn and nothing more.

Two practical traps

• & in links. Links pulled out of HTML or RSS often have & (sometimes &) between parameters. Replace it with & before you parse or send the link.
• No offline conversion. A short code cannot be turned into mid/idx/sn without fetching the article. Fetch once, store the long form or its sn, and dedupe on that.

How the API handles both

Official Account article detail takes either form as url. It returns the full text, title, account and publish time, and a url that is normally the long form. The article's id is its sn when the link has one, otherwise __biz, mid and idx joined with underscores, so the same article normally gets the same id whichever link you started from. Article history returns long links, with each article's idx as ext.position.

Task	Calls	Cost
One short or long link to full text	1	$0.02
A sheet of 400 mixed links	400	$8.00
1,000 links	1,000	$20.00

Example: clean, fetch, dedupe

import html, os, requests
from urllib.parse import urlparse, parse_qs, urlencode

def clean(link):
    link = html.unescape(html.unescape(link))          # & -> &
    q = parse_qs(urlparse(link).query)
    keep = {k: q[k][0] for k in ("__biz", "mid", "idx", "sn") if k in q}
    return "https://mp.weixin.qq.com/s?" + urlencode(keep) if len(keep) == 4 else link

seen = {}
for link in ["https://mp.weixin.qq.com/s/k3Vq8ZtYwLr2PxN0aHd5Fg",
             "http://mp.weixin.qq.com/s?__biz=MzA5NTI3MjIwMA==&mid=2650412873&idx=2&sn=8c1f0e5a3d9b47e6a2c4f17b0d9e3a65&chksm=bc2f"]:
    r = requests.get("https://api.okdata.online/v1/wechat_mp/articles/get",
                     headers={"Authorization": f"Bearer {os.environ['OKDATA_API_KEY']}"},
                     params={"url": clean(link), "fields": "id,url,title,published_at"}, timeout=60)
    item = r.json()["data"]["item"]
    seen.setdefault(item["id"], item)
print(len(seen), "unique articles")

Limits

• Deleted articles, and articles removed after a complaint, return not_found, which is not charged.
• Read and like counts are not part of article detail; metrics fields are null and meta.unavailable says why.
• More on pulling text out of article links in full text from a mp.weixin.qq.com link.