LangChain4j

Java gets a first-class LLM toolkit in LangChain4j, an independent open source project that adapts ideas from LangChain and Haystack into idiomatic APIs for the JVM rather than a direct port of Python code. A unified abstraction covers popular model providers and vector stores, so swapping OpenAI for Anthropic, Gemini, or a local Ollama model means changing configuration, not rewriting logic, and the same applies across embedding stores. Higher-level building blocks include AiServices, which turns an annotated Java interface into a working LLM-backed service, plus tool calling with MCP support, agents, chat memory, structured outputs, and RAG utilities spanning document ingestion through retrieval. First-class Spring Boot and Quarkus integrations fit it into standard enterprise stacks. Licensed Apache-2.0 with no commercial tier, the project has grown to about 12,700 GitHub stars and a large contributor base, making it the default choice for teams adding LLM features, agents, or retrieval pipelines to existing Java services without leaving their ecosystem.