Modular RAG Architectures with Java and Spring AI
summary
A working Spring Boot RAG demo uses Spring AI plus Vaadin UI, a local Llama/Mistral inference service, PostgreSQL with PGVector-enabled Postgres and Grafana/OpenTelemetry for dev-time observability; an Aronia starter boots database, telemetry and testcontainers. Documents flow through query transformers (translation + rewrite), token splitting, embeddings (nomic‑embed), a vector‑store retrieval, post‑retrieval compression, and a 3‑way query expander; an agent uses tool calls or MCP for web search. Live demo answers queries correctly, fails on Danish until translation is applied.
AI-written summary · about
chapters (20)
- 0:00Modular RAG with Spring AI
- 1:08LLMs need context
- 3:14Spring Boot RAG app
- 6:25Retrieval augmented generation explained
- 9:13Naive sequential RAG
- 12:46Search engine retrieval demo
- 13:31RAG with OpenTelemetry
- 16:41Data for vector stores
- 19:34Semantic search with vectors
- 23:12Vector store retrieval demo
- 26:04Multilingual RAG queries
- 27:51Query optimization
- 30:20Document compression
- 33:32Advanced RAG expansion
- 37:27Multi-query expansion demo
- 38:52Conditional RAG routing
- 40:20Agentic RAG tools
- 44:31MCP servers in RAG
- 45:57Spring Data vector search
- 48:27Spring AI resources
related talks
45mFrom RAG to Enterprise AI Agents: Building Intelligent Java Apps
01h 00mSpring AI
49mBuilding AI Agents with Spring & MCP
49mPractical AI in Java
48mCrafting your own RAG system: Leveraging 30+ LLMs for enhanced performance
41mFrom Chat to RAG to MCP: Enhancing Java Applications with AI
45mFrom ChatGPT User to RAG Implementer: A Developer's Journey
01h 03mBuilding AI powered Applications with Java