HypeSpam

HypeSpam v2.0 — Changelog

🐛 Bug Fixes

Fixed a major false-positive bug in the word filter. Previously, blocked words were matched using plain substring search, so a short blocked word like "oc" would also trigger inside completely innocent words — for example, typing "bedrock" got auto-muted because it contains "oc". Words are now matched as whole words only, using Unicode-aware boundaries (also fixes matching for Turkish characters). The same bug existed in the leetspeak/repeated-character variation check and has been fixed there too.

✨ New Features

whitelist-words config option — an extra safety net list of words that will never be flagged, even if they'd otherwise match a filter. /antispam test <message> — new admin command to test whether a message would get blocked, without needing a real player to trigger it. Near-duplicate spam detection (check-similarity) — catches spam that isn't 100% identical but very close (e.g. "hey" → "heyy" → "heyyy"), using a configurable similarity threshold. /antispam version command. Short-word warning on /antispam addword for words 3 characters or shorter.

🔧 Internal

Word-filtering logic is now centralized in a single shared method, so live chat filtering and the new /antispam test command always produce identical results.

⚠️ Config note: if you're upgrading from v1.0, add the new keys (blocked-words-whole-word-only, whitelist-words, check-similarity, similarity-threshold, similarity-min-length) to your existing config.yml, or delete it to regenerate the default.