Description
The Omni-Moderation Plugin is a chat moderation tool for Minecraft servers. It intercepts player chat messages, checks them using OpenAI's Omni Moderation API, and either blocks or allows the content. All messages are logged, and repeated content is cached to reduce API usage.
The plugin uses my Omni-Moderation Java Maven Library for seamless API integration, request handling, and moderation result parsing. This external library cleanly separates API logic from plugin functionality.

How It Works
Unlike traditional filters that block individual keywords, Omni-Moderation moderates the entire chat message in context.

• It does not block words directly, but rather evaluates the meaning and intent of the full message.
• This allows the AI to distinguish between truly harmful content and acceptable expressions that might include sensitive terms.
• For example, it can detect toxic intent even if the message is disguised as N1**** instead of Ni****, while allowing non-offensive uses like I love all Ni****.

This makes the system harder to bypass using common tricks like character substitution, but also introduces some limitations (see Disclaimer).

All chat messages are logged, allowing you to review what was allowed or blocked. This logging also makes it possible to monitor users who frequently trigger the moderation system. These users can then be manually punished if necessary, based on server policy.

The plugin only logs and monitors flagged messages. What you do with the logs is entirely up to you.

Features

• AI moderation via OpenAI
• Chat message logging (blocked and allowed)
• SQLite cache with compressed database storage
• Customizable moderation profiles
• Whitelist/Blacklist commands for manual overrides
• Message logging enables detection of suspicious users

Requirements

• Java 21
• Minecraft 1.21 or higher
• A valid OpenAI API key (required, but free to use under rate limits)

To use the plugin, you must sign up at OpenAI and create an API key. After signing in, you can find your:

• API key in your API keys page
• Project ID in your organization projects page
• Organization ID in your organization settings page

Moderation Profiles
Select how strictly content is flagged:

• FULL – Flag anything suspicious
• HIGH – 35%+ flagged
• MEDIUM – 55%+
• LOW – 95%+
• MINIMAL – 99%+
• NONE – Only monitors, does not block

Tip: After using NONE, reset the database to avoid whitelisting undesired messages.

Commands
Only accessible by moderators:

• /moderation status – Show plugin status
• /moderation whitelist <content> – Whitelist specific text
• /moderation blacklist <content> – Blacklist specific text
• /moderation reset – Clear the moderation database
• /configure profile <FULL|...> – Set moderation profile
• /configure apikey|projectid|organizationid – View current setting
• /configure apikey|projectid|organizationid <value> – Set credentials

Disclaimer
Omni-Moderation is a solid first line of defense for chat moderation, but it is not a complete replacement for human review or broader server moderation policies.

• It relies on OpenAI’s context-aware moderation model.
• This model can sometimes allow positively framed but inappropriate messages like I love all Ni****, which are not flagged.
• At the same time, it is robust against obfuscation tricks like N1****, recognizing toxic intent even when the text is masked.
• These limitations are rooted in the OpenAI moderation model and cannot be fixed by this plugin.

I recommend combining this with:

• A manual review system
• Server rules enforcement
• Ongoing updates to whitelists and blacklists