header

Plugin to Create AI Customized Join & Quit Messages

LLMNotifications generates funny, customised join and quit messages for players using AI.

It can use:

a local AI running on your server (Ollama), or
cloud AI such as OpenAI or Anthropic.

Messages can include player and server information such as name, world, playtime and more.

Example prompt:

Write a funny 4-line rhyme to welcome @player to the server. Mention their world (@world).

You can also store extra information about players using the player memory feature, such as hobbies or favourite music, allowing the AI to personalise messages even more.

Quick Start

Put the plugin jar in your server plugins folder.
Start the server once.
Open plugins/MinecraftLLMNotifications/config.yml.
Choose an AI provider and model.
Edit the prompts.
Run /llmnotifyreload.

Features

AI generated messages for:
- first join
- returning join
- player quit
Multiple prompts per event (chosen randomly)
Works with local AI or cloud AI
Placeholder support for player and server information
Safe fallback messages if the AI fails
Optional player memory system for personalised messages

Configuration

When the plugin first runs it creates:

plugins/MinecraftLLMNotifications/config.yml

In this file you can:

choose your AI provider
select the model
customise prompts
configure fallback messages
enable or disable player memory

Look through the file for comments explaining each option.

Example Prompts

First Join

Welcome @player to the server with a short funny rhyme.

Join

Write a sarcastic welcome back message for @player.

Quit

Write a dramatic goodbye message for @player leaving @world.

Keeping messages short usually works best for chat.

Player Memory

You can store extra information about players that the AI can use when generating messages.

Examples:

favourite music
hobbies
running jokes on the server

Example:

/llmnotifymemory add Steve loves cheese

The AI will automatically see this information when generating messages.

Supported Prompt Placeholders

You can use these in prompts and fallback messages:

@player
@display_name
@uuid
@world
@x, @y, @z
@health
@food
@level
@gamemode
@online_count
@is_op
@first_join
@last_played
@playtime_hours
@player_memory

Tips for Better Prompts

Keep output short so it fits in chat.
Ask for a specific number of lines if needed.
Describe the tone you want (funny, sarcastic, dramatic, etc).

Example:

Write a short medieval themed welcome message for @player in 3 lines.

Commands

Command	Description
`/llmnotifyreload`	Reload the plugin configuration
`/llmnotifymemory list <player>`	Show stored memory for a player
`/llmnotifymemory add <player> <item>`	Add information about a player
`/llmnotifymemory remove <player> <item>`	Remove a memory item
`/llmnotifymemory clear <player>`	Clear a player's memory
`/llmnotifymemory clearall`	Clear memory for all players

Ollama Help

If you choose to use your own locally hosted ollama server (you will need to know how to set this up). There are some Ollama specific settings in the config file that is created when the plugin first loads:

  ollama:
    base-url: "http://127.0.0.1:11434"
    # there are some great small models designed for humour and sarcasm
    model: "MeaTLoTioN/LisaBot"
    # Optional: how long Ollama should keep the model loaded (examples: "5m", "30m", "1h", "0")
    keep-alive: "1h"
    # Optional: periodically ping Ollama with a tiny generation request to keep model warm
    keepalive-ping-enabled: true
    # Frequency for keepalive ping task when enabled
    keepalive-ping-interval-seconds: 300

The keep alives stop your Ollama model from unloading as the initial load time for some models may be slow enough that the plugin times out. You can also increase the request-timeout-seconds setting in the config file.