What’s New in Spring AI 1.0 GA? A Deep Dive into Java AI Features

Spring AI 1.0 GA is officially released, offering a rich set of features that every Java developer should know.

The core capabilities include a ChatClient interface supporting 20 AI models (e.g., Anthropic, ZhiPu, DeepSeek, MiniMax) with multimodal I/O and JSON responses, a retrieval module compatible with 20 vector databases, sliding‑window and vector‑search chat memory, an @Tool annotation for tool calling, model evaluation, observability, and Model Context Protocol (MCP) support.

1. ChatClient Core Interface

Primary interface for interacting with AI models, supporting 20 models and multimodal input/output where available.

Comparison of model-specific features is provided.

2. Retrieval‑Augmented Generation (RAG) and Vector Stores

Vector store abstraction that adapts to 20 vector databases (e.g., Azure Cosmos DB, Weaviate) with SQL‑like filtering.

ETL framework using a pluggable DocumentReader to ingest files, web pages, GitHub, cloud storage (AWS S3, Azure Blob) and databases, with built‑in chunking, metadata enrichment, and embedding generation.

RAG pipeline built on QuestionAnswerAdvisor and RetrievalAugmentationAdvisor components, with a reference case “Retrieval Augmented Generation (R.A.G.) with Vector Stores”.

3. Chat Memory

Basic implementation MessageWindowChatMemory stores the most recent N messages using a sliding window and supports JDBC, Cassandra, Neo4j persistence.

Advanced VectorStoreChatMemoryAdvisor retrieves semantically similar past messages via vector search (see “Guide to chat memory implementation”).

4. Tool Calling

Declare tool methods with the @Tool annotation, enabling dynamic Bean registration or programmatic creation to call external functions such as weather queries or database operations (see “Guide to local tool calling”).

@Component
class DogAdoptionScheduler {
    @Tool(description = "schedule an appointment to pickup or adopt a dog from a Pooch Palace location")
    String schedule(int dogId, String dogName) {
        System.out.println("Scheduling adoption for dog " + dogName);
        return Instant.now()
                .plus(3, ChronoUnit.DAYS)
                .toString();
    }
}

5. Model Evaluation and Observability

Evaluation components: RelevancyEvaluator checks response relevance; FactCheckingEvaluator validates factual accuracy against context.

Observability integrates Micrometer to trace model latency, token usage, tool calls, and supports logging and Micrometer Tracing.

6. Model Context Protocol (MCP)

Client starter spring-ai-starter-mcp-client connects to MCP servers via stdio or HTTP‑SSE, with an example for Brave search.

Server starter spring-ai-starter-mcp-server combined with @Tool builds MCP servers, integrates Spring Batch/Cloud Config, and supports OAuth security.

7. Agent Support

Workflow‑driven agents provide evaluation, routing, orchestration, chaining, and parallel execution patterns.

Autonomous agents discover tools dynamically via MCP, maintain execution memory, and support recursive strategy optimization (Spring MCP Agent demo).

In summary, the Spring AI 1.0 GA release represents a major upgrade for Java developers, delivering powerful AI capabilities, extensive tooling, and frameworks that simplify building AI applications.