Building a Johan Liebert Discord Bot with Gemini AI

Introduction

As a fan of Monster and the character Johan Liebert, I wanted to create something unique—a Discord bot that responds with the calm, intelligent, and unsettling personality of Johan himself.

Instead of creating a standard chatbot, I decided to combine Discord.js with Google's Gemini API to build an AI-powered character bot.In this article, I'll walk through the project, the tech stack, challenges I faced, and what I learned along the way.

What is Johan Liebert?

Johan Liebert is the main antagonist of the anime and manga Monster. He is known for his intelligence, psychological manipulation, and mysterious personality.The goal of this project was not to recreate Johan perfectly, but to capture some of his speaking style and presence inside a Discord server.

Project Overview

The bot:

Listens to messages in Discord
Sends user prompts to Gemini AI
Generates responses inspired by Johan's personality
Returns responses directly inside Discord channels

Tech Stack:

Node.js
Discord.js
Google Gemini API
dotenv

Setting Up the Project

First, I created a Discord bot application and obtained a bot token.

Then I initialized a Node.js project:

npm init -y npm install discord.js dotenv @google/genai

Environment variables were stored in a .env file:

Connecting Discord.js

The bot uses Discord.js to connect to Discord and listen for incoming messages.

const { Client, GatewayIntentBits } = require("discord.js");

Once connected, the bot waits for messages and processes user input.

Integrating Gemini AI

The most interesting part was connecting Gemini.

Instead of manually writing hundreds of responses, I used Gemini to generate contextual replies.

This allowed the bot to:

Handle different conversations
Generate unique responses
Stay in character

Challenges I Faced

1. API Configuration Issues

Initially, I had trouble getting Gemini API requests working correctly.

I had to verify:

API keys
Environment variables
Request formats

2. Prompt Engineering

Making the bot sound like Johan was harder than expected. Generic prompts produced generic responses. I improved results by providing stronger personality instructions and behavioral constraints.

3. Discord Message Handling

Managing message events and avoiding unwanted bot responses required careful filtering.

What I Learned

This project taught me:

How Discord bots work internally
API integration in Node.js
Environment variable management
Prompt engineering basics
Debugging third-party services

Most importantly, I learned that building projects teaches far more than watching tutorials.

Future Improvements

Planned features include:

Memory system for conversations
Slash commands
Personality modes
Logging and analytics
Docker deployment
Hosting on a cloud server

Conclusion

Building the Johan Liebert Discord Bot was one of my most enjoyable projects so far.

It combined anime, AI, backend development, and Discord automation into a single project.

If you're learning backend development, I highly recommend building something based on your own interests. The motivation to keep improving the project becomes much stronger when you're genuinely excited about the idea.

Thanks for reading!

Command Palette