Spring AI Essentials

Before jumping into specific examples, here's the stuff that shows up in every Spring AI project.

ChatClient

Everything starts with ChatClient. Think of it like RestTemplate but for LLMs.

@Bean
ChatClient chatClient(ChatClient.Builder builder) {
    return builder
        .defaultSystem("You are a helpful assistant.")
        .build();
}

Inject it, call .prompt(), get a response. That's the whole pattern.

How every call works

Same shape every time:

Build a prompt (system message + user message)
Add advisors if you need them (memory, RAG, logging)
Call the model — .call() for sync, .stream() for streaming
Get the response — string, entity, or structured type

String answer = chatClient.prompt()
    .user("What is Spring AI?")
    .call()
    .content();

The key pieces

Thing	What it does
`ChatClient`	Sends prompts, gets responses
`@Tool`	Lets the LLM call your Java methods
`ChatMemory`	Remembers previous messages
`Advisor`	Middleware for requests/responses
`.entity()`	Maps LLM output to Java objects

Each of these has its own page:

Prompts — system prompts, guardrails, prompt catalogs
Tools — function calling with @Tool and ToolContext
Chat Memory — conversation memory with MongoDB
Structured Output — mapping responses to Java records
RAG — retrieval-augmented generation