Skip to content

v0.2.0 — Offline AI, Asset Pipeline, Dungeon Prefab

Choose a tag to compare

@MolochDaGod MolochDaGod released this 28 May 21:33

Grudge GameForge v0.2.0

Ollama Offline AI

  • New local AI provider — connects directly to localhost:11434
  • 4 models: Qwen 2.5 Coder 7B, Llama 3.2, DeepSeek Coder v2, Code Llama
  • 1-click setup scripts (Windows + Mac/Linux) — installs Ollama, pulls models, starts editor

Browser Asset Conversion

  • FBX/OBJ/STL to GLB via assimpjs WASM in-browser (no desktop app needed)
  • ZIP extract + auto-convert contained 3D files
  • PNG/JPG/WebP/JSON direct upload to R2

Dungeon Interior Prefab

  • 822-mesh dungeon scene (forge-scene.glb) as builtin template
  • Layered parent-child hierarchy: Structure / Props / Lighting / Spawns / Collision
  • Dark ambient environment with 6 torch lights

AI System Upgrades

  • MAX_TOKENS_CAP: 8192 to 15360 (server)
  • MAX_TURNS: 8 to 15 (client) with cooldown at turns 14-15
  • 13 AI models across 3 providers (Anthropic + Puter + Ollama)

Animation Library

  • 22-clip catalog (locomotion, combat, emote, utility)
  • Mixamo URL patterns for runtime resolution

Deployment Stack

  • Vercel (SPA) + Railway (API) + Cloudflare (R2/DNS/Workers AI)
  • Offline/online/hybrid deployment modes
  • Cross-platform dev script fix, BakeProgressToasts test fix

Created by Racalvin The Pirate King at Grudge Studio.