v0.1.0

First release.

Added

• Three modes over one request per account. --mode profile (one row
  per account: subscribers, bio, category, website, ids, what the page
  carries), --mode spotlight (one row per Spotlight video, with view,
  share, comment, boost and recommend counts), --mode story (one row per
  snap in the live story and every curated highlight). All three read the
  same server-rendered __NEXT_DATA__, so they cannot disagree about an
  account.
• Plain HTTPS by default, a browser as fallback. Measured 2026-09-24
  from a datacentre address: every profile page served to plain curl, 30
  of 30 in a row with no delay; 1.29 s over HTTP against 3.14 s through a
  browser for three accounts, identical rows. --transport auto starts a
  browser only if the site refuses.
• Playwright, Selenium and Puppeteer engines, plus the 2Captcha Scraper
  API client, all run live on 2026-09-24 and producing identical rows.
  The Scraper API cost $0.0005 per page.
• The traps this site sets, each handled and pinned by a check on a real
  capture: a hidden subscriber count published as "0" (written as
  null, told apart by the page's own JSON-LD); empty Spotlight slots
  carrying viewCount: "0" (not rows; counted in the sidecar);
  highlight snaps with no id of their own (keyed on the CDN content id);
  an ordinary account served as a username and nothing else (a row with
  public_profile: false, not a failure); Snapchat's own boilerplate
  titles and descriptions (dropped); and the not-found page, which must
  classify the same with no HTTP status (Selenium has none).
• A sidecar that says when the page is not the whole account —
  handles_with_more_than_page — because a profile lists about 25
  Spotlight videos and this repo does not follow the cursor beyond them.
• diff_runs.py with tracked columns per mode.
• A daily canary with no secrets, green from a bare GitHub runner on its
  first dispatch.

Not implemented

• Discover, Lenses, Snap Map, topic pages, Spotlight comments, and the
  cursor pages beyond a profile's first ~25 Spotlight videos.

Fixed in the core this repo was built from

Found while porting the family core, and fixed here:

• The Scraper API client recorded every run as mode: "video" and
  carried a sibling's unused ytInitialData reader and /player column
  list. It now records the mode actually run, which is what lets
  diff_runs.py compare a Scraper API run with an engine run.
• The credential scan's site exemption is gone. The core forgave
  32-hex strings inside its previous site's CDN URLs; nothing here needs
  that, the fixtures carry none, and the strictest rule now covers the
  largest files.