You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit ea02b60
Browse filesBrowse the repository at this point in the historyBrowse files
docs: correct README and data README claims that drifted from the code
The README said the hot list refreshes on a schedule and the model is
trained weekly. Both are regenerated by hand. It also said the full database
syncs weekly (SyncWorker runs every 6 hours), that the model threshold is 0.7
(the shipped weights carry 0.653332 and SpamMLScorer reads it from the file),
that a not-spam report subtracts votes (merge_community_reports only parks a
community-only row for review until the maintainer approves it), and that hot
ranges need 3+ numbers (generate_hot_list wants 4 qualifying numbers and 6
reporters between them).
data/README said reports arrive through Issues and PRs, that the app fetches
spam_numbers.json whole, and that PhoneBlock's hashed lookup runs at runtime.
Reports come through the Worker, the app diffs shards against the signed
manifest, and ExternalLookup has no PhoneBlock code left.
docs/reproducible-builds.md described CI APK artifacts no workflow builds.
CHANGELOG: d522387 dropped the v1.7.36 heading, so its notes sat under
v1.7.37. The Unreleased drain entry now names database versions 40 and 41,
since the 280 files went in over two merges (132 added, 28 + 3 updated).
Also clears the prose semicolons and spaced hyphens left outside the version
highlights, which the getting-started item moves out of the README.
0 commit comments