What's New in 4.0

What’s New in 4.0

AI / LLM Platform

• @AiEndpoint + pluggable AiSupport SPI — auto-detects Spring AI, LangChain4j, Google ADK, or Embabel from the classpath. Conversation memory, RAG interceptors, cost/latency routing, and fan-out streaming are built in.
• @AiTool framework-agnostic tool calling — declare tools once with @AiTool/@Param, portable across all AI backends via automatic bridge adapters (SpringAiToolBridge, LangChain4jToolBridge, AdkToolBridge).
• Per-endpoint model override — @AiEndpoint(model = "gpt-4o") selects a model per endpoint without changing global config.
• Multi-backend routing — @AiEndpoint(fallbackStrategy = FAILOVER) wires DefaultModelRouter for failover, round-robin, or content-based routing across multiple AI backends.
• AiMetrics SPI — MetricsCapturingSession records first-streaming-text latency, streaming text usage, and errors via a pluggable AiMetrics interface (ServiceLoader-discovered).
• Auto-detect ConversationPersistence — conversationMemory = true discovers Redis or SQLite backends via ServiceLoader; falls back to in-memory.
• Broadcast filter auto-registration — @AiEndpoint(filters = {CostMeteringFilter.class}) wires filters without manual Broadcaster config.
• Cache replay coalescing — reconnecting clients receive coalesced missed streaming texts in a single batch.
• Streaming text budget management — StreamingTextBudgetManager enforces per-session and per-endpoint streaming text limits.
• Cost metering UI — per-message cost badges in the chat frontend via StreamingSession request attribute.

MCP (Model Context Protocol)

• MCP server — expose tools, resources, and prompt templates to AI agents with @McpServer / @McpTool annotations over Streamable HTTP, WebSocket, or SSE.
• OpenTelemetry tracing — McpTracing auto-instruments tool/resource/prompt calls with span propagation.

Core Transport

• gRPC transport — bidirectional streaming alongside WebSocket/SSE/Long-Polling on the same Broadcaster.
• Virtual threads — enabled by default on JDK 21+. ReentrantLock replaces synchronized to avoid pinning.
• Rooms & presence — RoomManager with join/leave, presence events, message history, and AI virtual members.

Cloud & Persistence

• Clustering — Redis and Kafka broadcasters for multi-node deployments.
• Durable sessions — survive restarts with InMemory, SQLite, or Redis-backed session stores.

Integrations

• Spring Boot 4.0 / Quarkus 3.21 — first-class starters with auto-configuration, native image support, and observability (Micrometer + OpenTelemetry).
• Kotlin DSL — builder API and coroutine extensions for idiomatic Kotlin usage.

Client Libraries

• atmosphere.js 5.0 — TypeScript client with React, Vue, and Svelte hooks for chat, rooms, presence, and AI streaming.
• React Native / Expo — useAtmosphere React Native hook with EventSource polyfill, NetInfo injection, and markdown rendering for mobile AI chat. See the expo-client sample.
• wAsync 4.0 — Java client rewritten on java.net.http with gRPC support.

Samples — Forked & Augmented

Four official AI framework samples have been forked and augmented with Atmosphere’s real-time streaming, tool calling, cost metering, and content safety:

Sample	Forked from	Atmosphere features added
spring-boot-langchain4j-tools	langchain4j-examples/spring-boot-example	Tool calling, PII redaction, cost metering
spring-boot-spring-ai-routing	spring-ai-examples/routing-workflow	Prompt routing, content safety, cost metering
spring-boot-embabel-horoscope	embabel-agent-examples/horoscope	Step progress streaming, content safety
spring-boot-adk-tools	adk-java/city-time-weather	Tool calling, streaming text budgets, response caching

Developer Experience

• Architectural validation — CI gate detects NOOP/dead code, placeholder stubs, DI bypass, and fluent builder misuse via TOML-configured patterns.
• atmosphere-generator — CLI scaffolding with --tools flag to generate @AiTool methods.

See Also

• Full Documentation