New Users Anti Spammer System

Open Rocket.Chat communities face a growing problem where new accounts join, flood multiple public channels with spam, and disappear before any moderator can respond. Current moderation is entirely reactive and manual, which is simply too slow for coordinated spam campaigns. This project builds a proactive, human-validated anti-spam system that monitors new users during their first 6 to 10 weeks and responds through a graduated restriction system called Moderation Levels, where every action is proportional, reversible, and requires admin approval before anything permanent happens. The system is built as a Rocket.Chat App on the Apps Engine with a seven gate detection pipeline running entirely in memory for speed. Gates catch exact duplicates via hash matching, polymorphic spam via cosine similarity, cross channel floods, rate flooding, rapid room spread, and phishing link campaigns. A four tier Moderation Level system escalates naturally with repeat behavior, from a friendly DM warning through posting cool downs to a permanent block that only lifts when an admin vouches for the user. The admin control layer includes a live dashboard, a full slash command suite, and AI powered on demand summaries so admins have everything they need to make confident moderation decisions without opening a single log file. Deliverables include a fully functional installable Rocket.Chat App covering the complete detection pipeline, a configurable Moderation Level system, a behavioral scoring engine, enriched admin notifications with message history, AI powered user analysis commands, a paginated admin dashboard built for scale, daily scheduled reporting, and complete documentation covering architecture, configuration, and known limitations.