Thinking System anti-cheat intelligence engine for Paper Minecraft.
TS-Spear reframes anti-cheat from boolean flags into a state-driven confidence system. Every action produces evidence; evidence propagates through typed graphs; tension resolves into explainable confidence; staff receive recommendations — never silent autobans by default.
Part of the BoggersTheFish Thinking System (TS) ecosystem. Architectural discipline from BadgersMC/spear-plugin (SPEAR methodology).
Phase 5 — ML-Ready · Paper 1.21+ · Java 21
| Phase | Scope | Status |
|---|---|---|
| 0–1 | Foundation, MVP checks, SQLite, decay, replay | ✅ |
| 2 | Combat checks, GUI suite, JSON export | ✅ |
| 3 | Freeze, JDBC backends, i18n, check management | ✅ |
| 4 | Object pooling, packet checks, JMH benchmarks | ✅ |
| 5 | ML export, annotations, inference SPI | ✅ |
| 6 | Federation, dashboard, marketplace | Planned |
See Implementation Phases for full delivery detail.
./gradlew build
# Plugin JAR: plugin/build/libs/TSSpear-0.1.0-SNAPSHOT.jarCopy the JAR to plugins/, restart Paper, run /ts inspect <player>.
Full guide: Installation
| Document | Description |
|---|---|
| Documentation Index | Complete doc map |
| Installation | Deploy on Paper |
| Configuration | All config keys |
| Commands | Staff /ts reference |
| Storage | SQLite, MySQL, PostgreSQL |
| Development | Build, modules, testing |
| ML Hooks | Export format, inference SPI |
| Architecture Design | Full system design (1,300+ lines) |
| Core Interfaces | Port contracts |
| GUI Wireframes | Staff interfaces |
api/ core/ storage-sqlite/ storage-jdbc/
checks/movement/ checks/combat/ checks/networking/
benchmarks/ plugin/
| Command | Description |
|---|---|
/ts inspect [player] |
Confidence inspector GUI |
/ts export [jsonl] [player] |
ML-ready evidence export |
/ts annotate <player> <verdict> [notes] |
Training label |
/ts infer [player] |
Run inference provider |
/ts freeze [player] |
Continuous replay capture |
/ts debug profile |
Performance metrics |
Full reference: Commands
- Nothing is boolean — confidence is multi-dimensional and continuous
- Checks emit evidence, not punishments — staff own enforcement
- Everything is explainable — scores decompose into evidence
- Everything produces receipts — immutable confidence audit trail
- Everything is replayable — tick-by-tick suspicious sequences
- AI-ready by contract — stable ML export without architectural churn