Chat memory gets fuzzy fast once the UI hides what LangChain4j is actually retaining.

I wrote a Quarkus tutorial that makes retained-memory pressure visible with `TokenWindowChatMemory`, Ollama request counts, a turn ledger, and OpenTelemetry attributes. The useful split is simple: your app-level eviction budget is not the model context limit. https://www.the-main-thread.com/p/quarkus-langchain4j-chat-memory-budget #Java #Quarkus #LangChain4j #Ollama #OpenTelemetry

Local AI gets risky when the first confident answer becomes the system answer.

I wrote a Quarkus tutorial that sends the same text to two Ollama models, uses Quarkus Signals to escalate only on disagreement, and keeps `UNCERTAIN` separate from `FAILED`. https://www.the-main-thread.com/p/quarkus-langchain4j-ollama-signals #Java #Quarkus #LangChain4j #Ollama

Most #AI prototypes work. Until the next model update breaks half the system. Lutske de Leeuw & Maarten Vandeperre show how #CleanArchitecture, ports & adapters keep AI integrations from becoming spaghetti code.

Read: https://javapro.io/2026/03/17/ai-without-spaghetti-clean-architecture-in-the-age-of-ai/
#LangChain4j #Quarkus QuarkusIO #LLM

Our next #JCON2026 session is live: 'Talk to Your Data: Natural Language Data Access in #Java with #Hibernate #Quarkus and LangChain4j' with Marco Melladelli

Explore how Hibernate ORM, Quarkus, and #LangChain4j come together to enable …

Grab your coffee and hit play: https://youtu.be/tMW5jxX6DoA

Tired of stitching #AI SDKs into your #JakartaEE stack manually? #LangChain4J-CDI lets you declare an interface, annotate it, & inject it anywhere — REST, EJB, schedulers. @EliteGentleman demonstrates the model-driven approach.

Worth a closer look? Read: https://javapro.io/2026/02/25/bring-ai-into-your-jakarta-ee-apps-with-langchain4j-cdi-formerly-smallrye-llm/

Confused by the exploding number of #AI tools in the #JVM ecosystem? Teams mix #SpringAI, #LangChain4j, MCP & #Ollama without understanding the layers underneath. Artur Skowronski explains what each part of the #Java AI stack is actually for: https://javapro.io/2026/06/03/the-gen-ai-iceberg-java-tooling-edition/

@langchain4j

Cheap questions should not burn the same local model as real debugging work.

I wrote a Quarkus + LangChain4j tutorial that classifies prompts, routes them between two Ollama models, and keeps the decision observable with CDI events and tests. https://www.the-main-thread.com/p/quarkus-langchain4j-model-routing #Java #Quarkus #LangChain4j #Ollama

LangChain4j: Chat With Documents

#ai #java #langchain4j

https://mydeveloperplanet.com/2024/01/24/langchain4j-chat-with-documents/

LangSmith does not need to stay in the Python corner.

I wrote a Quarkus walkthrough that sends LangChain4j traces to LangSmith over OTLP, including plain chat, tool calls, and controlled failures. It also covers the two easy footguns: /otel vs /otel/v1/traces, and region-specific endpoints.

https://www.the-main-thread.com/p/quarkus-langchain4j-langsmith

#Quarkus #LangChain4j #OpenTelemetry #Observability