Building and Deploying A Multi-Agent AI D&D with Typescript
summary
Builds a working multi‑agent Dungeons & Dragons demo that actually creates characters, rolls dice, queries rules from a PDF, and runs a front end—code is TypeScript‑based and executed live. Uses trypescript with Strand Agents, an MCP/HTA orchestration layer, an Amazon Bedrock Haiku 4.5 model option, Express and streamable HTTP, a vector DB built from a 2018 rules PDF via Hugging Face transformers and a local mini‑LLM, and custom tools (roll_dice) that demonstrably work.
AI-written summary · about
chapters (14)
- 0:00Gaming and text adventures
- 3:31Zork's AI inspiration
- 4:56AI agents and lifecycle
- 6:20Strand Agents for TypeScript
- 7:23Basic agent with system prompts
- 9:50Adding HTTP and file tools
- 13:03Custom dice rolling tools
- 14:49Centralizing logic with MCP
- 18:01Connecting MCP client-server
- 19:25Multi-agent HTA architecture
- 21:31Rules and character agents
- 24:01Running multi-agent system
- 25:23Live D&D character demo
- 27:06Q&A
related talks
10mRapid Prototyping with AI
02h 48mCompose & Dragons: The Role-Playing Game of Agents Fed on Tiny Language Models
23mBuilding AI Agents in TypeScript
01h 40mOrchestrating a Role-Playing Game Universe with Multi-Container AI Agents
43mCoding a Multi-Agent Game Master with Spring AI
20mShip it! Building Production Ready Agents
31mReal-World Multi-Agent Development | Ship Faster with AI Agents
50mSpec-driven Development using Coding Agents