collinzrj

collinzrj / TokenSpire2

Public

A SlayTheSpire2 Agent framework to evaluate continual learning capabilities of LLM

15
3
100% credibility
Found Mar 22, 2026 at 15 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
C#
AI Summary

A mod for Slay the Spire 2 that automates full gameplay runs using AI to make decisions in combat, navigation, events, shops, and more, with built-in learning across multiple sessions.

How It Works

1
🕵️ Discover the smart game helper

You hear about TokenSpire2, a clever companion that plays Slay the Spire 2 all by itself on places like GitHub.

2
📥 Get it ready

Download the simple package and place it right into your game's special add-ons folder.

3
🧠 Link the thinking brain

Connect a smart AI helper service so it can make wise choices during the game.

4
▶️ Start auto-adventure mode

Launch your Slay the Spire 2 game with the easy auto-play setting turned on.

5
👀 Watch the magic unfold

Relax as it handles every moment—battles, maps, shops, events, and rewards—running full adventures on its own.

6
📈 Witness it grow smarter

After each game, it thinks back on what happened, learns from wins and losses, and gets better over time.

🏆 Epic victories unlocked

Your AI buddy climbs higher floors, defeats tough bosses, and masters the spire like a pro gamer.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 15 to 15 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is TokenSpire2?

TokenSpire2 is a C# mod for Slay the Spire 2 that turns the game into an autonomous agent framework, where LLMs make all decisions to play full runs. It serializes game states—like combat turns, maps, events, shops, and rewards—into text prompts sent to OpenRouter-compatible APIs, parsing responses into actions like "PLAY 3 -> A" or "CHOOSE 2". Developers get a ready-to-run setup to evaluate LLM capabilities in a complex, turn-based roguelike, complete with bilingual prompts, fallback random mode, and detailed logs.

Why is it gaining traction?

It stands out with a continual learning system: after each run, the LLM reflects on failures (e.g., lacking strength scaling) and updates a persistent memory file carried across sessions, showing real improvement like reaching Act 2 bosses. Benchmarks compare models—Claude Opus leads at 17.2 average floors on Ironclad A0—while handling every game phase without babysitting. The hook is stress-testing agent frameworks in a high-stakes environment that demands strategy over simple tasks.

Who should use this?

AI researchers benchmarking LLM agents for decision-making and continual learning in games. Slay the Spire 2 modders integrating LLM-driven bots into custom scenarios. Devs prototyping C# tools for OpenRouter APIs with built-in game automation and JSON logs for analysis.

Verdict

Grab it if you're experimenting with LLM agents—solid docs, performance charts, and easy setup via Steam launch args make it immediately usable despite 15 stars and 1.0% credibility signaling early maturity. Polish tests and expand model support to boost adoption.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.