Bringing stories to life with AI, data streaming and generative agents
summary
Prototype builds a mini Stardew Valley village where characters walk, start dialogue when within ~1,024 pixels, and exchange memories; position events stream to Apache Kafka, distances and joins are computed with Apache Flink, replies are generated by an OpenAI LLM, and conversation summaries become embeddings stored in a vector search/RAG index. Demo runs with a small cast (three-to-twenty agents), limits conversations to two participants, requires prompt tuning and cooldowns, and successfully seeds a propagated goal (a farming “conference”) though memory summarization needed to be richer.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:24AI agents and simulations
- 10:18Stanford village simulation
- 20:45StarDual village demo
- 23:35Apache Kafka for real-time
- 25:20Apache Flink for conversations
- 28:15LLM character dialogues
- 32:33Implementing memory with retrieval augmented generation
- 36:05Agent goals and ideas
- 39:34Future social science plans
- 40:36Q&A
related talks
43mRoleplAI
01h 34mHands-on AI workshop: Graph RAG, Memory & Multimodal Agents
24mBuilding a Viral Open Source AI Chatbot: Journey from Concept to Reality
54mGenAI Grows Up: Building Production-Ready Agents on the JVM
29mTools and Roadmap for AI-Native Developers
51mEvent Driven AI: Supercharging ChatGPT with RAG & LangStream
34mBuilding Multi-Agent Generative-AI Applications with AutoGen
01h 07mBuilding and Scaling an AI Agent Swarm of low latency real time voice bots