ChatSentinel vs PunishmentSystem

ChatSentinel is a self-hosted moderation plugin for Minecraft Paper servers. Its local trie filter checks chat and can load word lists for 28 languages, with leet-speak normalization and abbreviation expansion. It can also moderate whispers, signs, books, and anvil renames. Flagged content can be blocked or masked, with configurable escalation to warnings, mutes, or staff alerts. Optional moderation API and LLM review layers can add checks; both are disabled by default, so the local filter can operate without API keys. The first layer uses an OpenAI-compatible moderation endpoint. The second accepts compatible chat-completions providers, including Ollama, LM Studio, vLLM, Azure OpenAI, Together, and Groq. SQLite is the default database, with MySQL and PostgreSQL also supported. An optional PlaceholderAPI integration exposes moderation-related player details as placeholders. The README lists Java 21 or later and a Paper server as requirements. It describes a fail-open pipeline: local filtering remains active if APIs are unreachable, with a configurable timeout before messages pass through. ChatSentinel is free and open source.

PunishmentSystem is a free, open-source Minecraft server plugin for administering player discipline through a graphical interface. It supports bans, temporary bans, mutes, kicks, and unbans, while maintaining searchable punishment histories with pagination, filters, and export options. Release 1.1.0 lists a database, case and points systems, a case-lookup plugin, and an API. Discord notifications can include the punishment type, reason, staff member, duration, timestamp, player details, and server name. Ban-evasion detection tracks IP addresses, compares banned players with new joiners, alerts staff with the required permission, and records suspected attempts. The README lists EssentialsX integration and support for Paper, Spigot, and Folia servers running Minecraft 1.21 or later with Java 11 or later. Bedrock Edition players can use a configurable prefix. The project says punishment data uses AES-256 encryption, secure key handling, tamper protection, and automatic backups. It is self-hosted and licensed under GPL v3.