Developing Cloud-Native Java AI applications with DJL and LangChain4J
summary
Shows concrete Java pipelines: uses DJL to train or infer (example: an 8‑megabyte ONNX fraud model that maps float inputs to boolean decisions) and LangChain4J + Quarkus to wire LLMs, embeddings and tools. Demos include retraining patterns (nightly refeeds), RAG with three transcripts chunked to ~500 tokens, vector stores (PGVector / Chroma) with 384‑dimension embeddings and five nearest‑neighbor recommendations, Kafka/Redis memory, J Llama/AMA local models, LangGraph4J flows, and OpenTelemetry/Grafana tracing; most demos build and run, some live demos are flaky.
AI-written summary · about
chapters (12)
- 0:00Introduction and context
- 3:58AI concepts and DJL
- 11:20Model inferencing demo
- 22:18Introduction to LangChain4j
- 26:58Prompting and memory demos
- 37:01Advanced LangChain4j integration
- 56:14RAG and Tooling implementation
- 1:08:46RAG with vector databases
- 1:36:47Embedding-based recommendations
- 1:47:23Security with Guardrails
- 1:53:32LangGraph and observability
- 2:18:56Closing thoughts and demo finale
related talks
01h 01mScaling AI with Java: Building Production-Ready LLM Applications
01h 01mCreate #AI Infused #Java Apps with #LangChain4j
53mJava meets AI: Build LLM-Powered Apps with LangChain4j
47mJava Meets AI: Integrating LLMs into Your Applications with LangChain4j
50mDeveloping AI applications in Java with Quarkus and Langchain4j
01h 03mBuilding AI powered Applications with Java
47mQuarkus and AI Integrating Java and LLM to build intelligent applications
48mBuilding AI Powered Applications with Java