Skip to main content

​
Thaumcraft 4 Research Bot

A screenshot-based bot that automates the Thaumcraft 4 research minigame for Minecraft 1.7.10, designed to replace manual research helpers and websites.

​
Overview

This bot uses computer vision to recognize research puzzles directly from your game screen, solves them using an efficient algorithm, and automatically places aspects using mouse control. Built specifically for the Gregtech: New Horizons modpack with TC Research Tweaks addon support.
The bot has been tested to work on all research puzzles in GTNH and generates solutions optimized for minimal aspect cost.

​
Key features

Pixel-based recognition

Recognizes puzzle boards directly from screenshots using custom resource pack markers

Universal solver

Fast, efficient algorithm that works on all research puzzles with optimized aspect usage

Automatic mouse control

Quickly inputs puzzle solutions and can craft undiscovered aspects automatically

Configurable costs

Customize aspect costs to prioritize abundant materials in your inventory

​
How it works

  1. Screenshot capture - Takes a screenshot of your game window with the research table open
  2. Puzzle parsing - Identifies the hex grid, existing aspects, and empty spaces using pixel detection
  3. Solution generation - Runs an optimized solver to find the cheapest path connecting all aspects
  4. Automatic placement - Uses mouse control to place aspects according to the solution
The bot generates solutions that prefer simple aspects (like primal aspects) to minimize material costs.

​
Test modes

The bot includes multiple testing modes for development and validation:
  • Test mode - Uses debug_input.png instead of taking screenshots, skips mouse actions
  • Test-all mode - Runs solver on all puzzles in test_inputs/ folder for benchmarking

Installation

Get started with uv and prerequisites

Quick start

Solve your first research puzzle

​
Platform support

Linux is not currently supported. The bot requires Windows for screenshot capture and mouse input functionality.

​
Performance notes

  • Optimized for speed in Python with efficient pathfinding algorithms
  • Handles most puzzles in seconds
  • Large boards with 7+ given aspects may take longer (potentially minutes for 9+ aspects)
  • Solutions are cached to improve performance on repeated patterns

Build docs developers (and LLMs) love

Get started for freeTalk to us