Donut Algorithm

This algorithm is made for the DONUTSMP auction house. It tracks arbitrage opportunities through previous data. This project was completely made with Ai, specifically GPT SOL 5.6

Credits: https://youtu.be/qK_7ByDtBbM?si=dSBCk3gnvKjUJ4mk

Inspired the basis for the algorithm and how machine learning can use previous statistics to find price weakness. 🙏

Install and use

1. Install Fabric Loader for Minecraft 1.21.11.
2. Put Fabric API 1.21.11 and the built JAR in Minecraft's mods folder.
3. Start Minecraft with Java 21.
4. Press Right Shift to open or close the ClickGUI. Escape closes it.

Enable live prices

1. In DonutSMP, use /api to create a key. (Only Applicable if you purchase DONUT+) 🥲
2. Close Minecraft.
3. Create .minecraft/config/donut-flip-scanner/api-key.txt and paste the key only. Do not include the word Bearer; normal editor-added whitespace is trimmed.
4. Start Minecraft again. The status bar changes from API: Mock Data to Connecting, then Connected after the first successful official API response.

The client immediately polls recently listed auctions, completed transactions, and the first active-listing page. It then checks recent listings every 5 seconds and completed transactions every 30 seconds. Transaction page 1 is sampled every other poll, with pages 2-10 overlapped for historical evidence, while active-listing pages 1-10 rotate every 2 minutes. Listing prices shown by live opportunity cards come directly from the official response. Fair value, profit, ROI, and confidence are locally calculated from completed-sale evidence, so those cards can take a short warm-up and will remain empty when the requirements are not met.

Local files

The persistence APIs use:

config/donut-flip-scanner/config.json
config/donut-flip-scanner/api-key.txt
config/donut-flip-scanner/market-data.db
config/donut-flip-scanner/exports/
config/donut-flip-scanner/backups/

See release hardening, API integration, database design, and GUI/backend integration.