/springone-2024-spring-ai

Primary LanguageJava

Bootiful Spring AI

by Dr. Mark Pollack, Dr. Rod Johnson, and Josh

you can find Rod's code here

https://chat.lmsys.org

  • "hi, Spring fans!"
  • DJ (~5 mins): officiate the project
    • JVM devs are important to the future of AI.
    • people are not going to switch to Python.
    • Adding GenAI to existing applications with Spring AI is trivial. Want to add functions to your AI flow, woohoo! You've got a lot of valuable code in Spring already, and it's trivial to connect the model to your APIs.
    • you should be excited about this!
    • requires a different skill set to make apps useful than model builders and data scientists
    • Spring's been around for a really, really long time.
    • Portable Service Abstractions, AOP, DI, autoconfiguration
  • DP (~5 mins):
    • history of AI
    • problems/solutions
    • constraints inform architecture (microservices vs. team size, models vs. speed/intelligence)
    • choose your model wisely (costs vs. speed/accuracy) "which one should we choose?"
    • "vibe checks are all you need."
    • https://chat.lmsys.org
  • JL/DP:
    • start.spring.io: web, graalvm, jdbc, openai, pgvector, actuator, zipkin, docker compose
    • setup :
      • API key
      • remove Docker Compose from pom.xml, specify spring.datasource.*, export the ports in compose.yml, docker compose up -d
      • model
      • Spring AI M2!
  • DP/JL: introduce the ChatClient with a simple joke controller
  • DP/JL: structured output (Joke.class)
  • DP/JL: RAG
  • DP/JL: Chat Memory
  • DP/JL: functions (show getting the status of a purchase (a Transaction))
  • DP/JL: observable AI: observability
  • DP/JL: efficient AI: virtual threads and GraalVM