Playing D&D With AI: Building an Agentic Game Engine
summary
A combat engine is built that pairs LLM reasoning with explicit code so the model suggests actions while deterministic functions enforce rules, state, and initiative; a 3‑day hackathon prototype grows into a two‑month build that reliably runs multi‑turn D&D fights across full turns (examples: thunderclap triggers eight tool calls to resolve saves and damage). The system uses domain classes, limited per‑turn context and a react thought‑action‑observation loop and runs on Infobip Agentic Workflows.
AI-written summary · about
chapters (10)
- 0:00Introduction to D&D and AI
- 3:39Challenges of multi-turn D&D combat
- 6:51Agentic systems workflows vs agents
- 9:38Context engineering for reliability
- 11:45Building an AI dungeon master
- 15:16REACT pattern in gameplay
- 18:22Graph structure and state management
- 19:43Complex spell interactions demo
- 20:25Info platform for agentic systems
- 23:36Summary and Q&A
related talks
46mThe Future of Role-Playing Games IS... A.I. | But How?
02h 48mCompose & Dragons: The Role-Playing Game of Agents Fed on Tiny Language Models
51mRolePl-AI
18mThink You Can Build a Game with AI? Think Again!
44mBuilding Agentic AI Systems: From Transactions To Conversations
34mCan AI Make Your Game Better?
29mAgentic AI: What It Is and Why It Matters
56mDesigners, Dungeons and Dragons