spring-boot-spring-ai-chat

Atmosphere + Spring AI Chat

Real-time AI chat using Spring AI's ChatClient with Atmosphere's SpringAiStreamingAdapter for WebSocket streaming.

What This Demonstrates

This sample shows the bridge between Spring AI and Atmosphere:

User → WebSocket → @ManagedService → SpringAiStreamingAdapter
                                        ↓
                                    ChatClient.prompt(msg).stream()
                                        ↓
                                    Flux<ChatResponse> → StreamingSession.send(streamingText)
                                        ↓
                                    WebSocket ← text-by-text to browser

Key point: You keep your Spring AI code (ChatClient, Advisors, tools) and get real-time WebSocket push for free via atmosphere-spring-ai.

Running

# From the repository root
cd samples/spring-boot-spring-ai-chat

# Without API key (demo mode — simulated streaming)
mvn spring-boot:run

# With OpenAI
export OPENAI_API_KEY=sk-...
mvn spring-boot:run

# With Gemini (OpenAI-compatible)
export OPENAI_API_KEY=AIza...
export SPRING_AI_OPENAI_BASE_URL=https://generativelanguage.googleapis.com/v1beta/openai
export SPRING_AI_OPENAI_CHAT_OPTIONS_MODEL=gemini-2.5-flash
mvn spring-boot:run

Open http://localhost:8083 in your browser.

Key Code

The Handler (SpringAiChat.java)

@Message
public void onMessage(String userMessage) {
    var session = StreamingSessions.start(resource);
    var chatClient = SpringBeanAccessor.getBean(ChatClient.Builder.class).build();
    var adapter = SpringBeanAccessor.getBean(SpringAiStreamingAdapter.class);
    adapter.stream(chatClient, userMessage, session);
}

Three lines to bridge Spring AI to real-time WebSocket streaming:

1. Create an Atmosphere StreamingSession
2. Get Spring AI's ChatClient
3. Stream via the SpringAiStreamingAdapter

Dependencies

<!-- Atmosphere with Spring AI adapter -->
<dependency>
    <groupId>org.atmosphere</groupId>
    <artifactId>atmosphere-spring-ai</artifactId>
</dependency>

<!-- Spring AI OpenAI starter (or any model provider) -->
<dependency>
    <groupId>org.springframework.ai</groupId>
    <artifactId>spring-ai-starter-model-openai</artifactId>
</dependency>

Architecture

Component	Role
atmosphere-spring-ai	Bridges Flux<ChatResponse> → StreamingSession
SpringAiStreamingAdapter	Auto-configured bean that does the bridging
ChatClient	Spring AI's fluent API for LLM interactions
StreamingSession	Atmosphere's text-by-text push to browsers

Environment Variables

Variable	Description	Default
OPENAI_API_KEY	API key for OpenAI or compatible provider	demo (demo mode)
SPRING_AI_OPENAI_BASE_URL	Override API endpoint (for Gemini, etc.)	OpenAI default
SPRING_AI_OPENAI_CHAT_OPTIONS_MODEL	Model name	gpt-4o-mini