Helidon 4.4.0: LangChain4j Integration and OpenTelemetry Modernisation
Helidon 4.4.0 is a feature and maintenance release for Oracle's microservices framework. The headline addition is experimental support for LangChain4j 1.11.0, bringing AI/LLM integration capabilities into Helidon applications without switching frameworks. The integration is marked experimental given the rapid pace of the LangChain4j API, but it gives Helidon developers access to the full LangChain4j model provider catalogue — including local models, OpenAI-compatible endpoints, and Anthropic's API — with Helidon's dependency injection and configuration system wiring the pieces together.
On the observability side, Helidon 4.4.0 deprecates Jaeger-specific tracing support in favour of the OpenTelemetry Collector approach, which is now the recommended path for trace export. Jaeger support has been superseded by the OTLP export protocol that the OpenTelemetry Collector accepts, making the Jaeger-specific module redundant. Teams relying on direct Jaeger export should plan a migration to the OTel Collector.
The release also ships JSON provider enhancements that improve interoperability with Jakarta JSON-P and Jakarta JSON-B, and includes the usual set of dependency version bumps to keep the framework up to date with the broader Jakarta EE and MicroProfile ecosystem.
Read more — InfoQ
Open Liberty 26.0.0.3 Beta: MCP Server Support for AI Agents
IBM's Open Liberty 26.0.0.3 beta introduces Model Context Protocol (MCP) Server support, allowing Open Liberty applications to expose themselves as MCP-compatible tools to AI agents and orchestrators. The implementation ships new interfaces that application code implements to define tools, resources, and prompts in the MCP vocabulary, and the Liberty runtime handles the protocol-level transport and discovery.
The MCP Server feature in this beta also introduces session-based request ID functionality. Each MCP session gets a stable identifier that persists across the client-server connection, which is important for stateful agent interactions where the agent needs to correlate multi-turn conversations or long-running tool invocations back to a single session context. This goes beyond simple request-response and supports the kind of persistent agent-application relationships that modern LLM orchestration frameworks expect.
Open Liberty 26.0.0.3 beta is the preview build ahead of the 26.0.0.3 GA release, which will align with the Jakarta EE 11 and MicroProfile 7 specifications. The MCP Server feature is the most forward-looking addition, reflecting the growing expectation that enterprise Java runtimes become direct participants in AI agent pipelines.
Read more — InfoQ
Java Operator SDK 5.3.0: Read-After-Write Consistency and Controller Metrics
Java Operator SDK (JOSDK) 5.3.0 focuses on correctness guarantees and observability improvements for teams building Kubernetes operators in Java. The release strengthens resource operation semantics to guarantee read-after-write consistency: after a write operation completes, a subsequent read of the same resource will always reflect the written state. This eliminates a class of reconciliation bugs where operators would read stale state immediately after updating a resource, causing spurious re-queues or incorrect status updates.
Version 5.3.0 also introduces MicrometerMetricsV2, a redesigned metrics integration for Micrometer. The V2 API provides controller-scoped metrics, meaning each reconciler in an operator gets its own set of metric tags rather than sharing a flat metric space. This makes it possible to build dashboards and alerts that are specific to a single controller — reconciliation rate, error rate, and latency per resource type — rather than having to filter aggregated metrics. The V1 Micrometer integration remains available but is now considered legacy.
The JOSDK project is widely used as the foundation for Quarkus Operator SDK and other Java operator frameworks. The 5.3.0 improvements address issues that have been reported in production environments running large numbers of custom resource instances where the consistency gap and metrics ambiguity were most visible.
Read more — InfoQ
Links & Sources
You May Also Enjoy
