GoogleGeminiAIGod

Behold! The AI God!

This is a free-tier Google Gemini based version of the ChatGPT minecraft roleplay bot with some additional upgrades.

Modified version of the ChatGPT bukkit plugin by "BigYous" which uses the google gemini (within free-tier limits) to provide an improved AI roleplay experience free of charge. Join a world ruled by the AI and hope that your new overlord is having a good day. You will be sent on missions from God, have your buildings and photos judged, be rewarded greatly, or explode spectacularly. Tread carefully....... it see's and hears everything....

Local Setup

• clone the repo
• run ./gradlew in the root of the repo
• download paper mc server version 1.21.1
• setup the server
• download voicechat bukkit plugin 2.5.24
• download imagery api 1.2.1
• download ImageFrame 1.7.10 for printing out photos on maps
• place the jar files in the plugins folder of your server
• install fabric for 1.21.1
• download the fabric api jar for extra mod compatibility
• download the voice chat mod version 1.21.1-2.5.24 for fabric
• place those in your mods folder in .minecraft
• (Optional) install iris + sodium and BSL Shaders for a more godly experience

Config

• Get a free gemini api key from aistudio.google.com/
• Get a free speechify api key from console.sws.speechify.com/tts
• configure primary model to be model-name: gemini-1.5-pro for smarter role-play output at slower speeds. Both models are fun, so give them both a try
• configure primary model to be model-name: gemini-1.5-flash if you want fast model times such as rate: 30 or below.

• NOTE: Do not go below 30-32 on the free tier pro model or you will hit your 2 requests per minute and 50 requests per day fast
• NOTE: if you set your primary model to gemini-1.5-flash then set secondary to gemini-1.5-flash-8b as free-tier rate limits are per-model name
• do not use the 1.0 models or other models which are not multimodal for secondary model. This program relies on the audio and visual processing of the latest Gemini 1.5 models.
• the flash model can go below 20 and performs quite nicely with the rapid feedback
• for a fast-feedback experience try setting both models to gemini-1.5-flash and try a rate of about rate: 20 to rate: 30
• do NOT use gemini-1.5-pro as the secondary model

To build the plugin for yourself from source see the instructions on the repository.

Running

• run the server once then go to plugins/gptgodmc/config.yml
• paste in your OpenAi API or Gemini key
• run the server again
• launch minecraft with fabric (make sure voice chat mod is installed)
• NOTE: on MacOS you need to use a launcher such as prisim to allow microphone access permission to for the plugin

• more info here

• finally, connect to the server at localhost