fit-framework

Why we included this project

Java teams have spent years watching the Python AI stack get all the good tooling, and FIT is the most complete attempt yet to bring that to the JVM. It is actually three pieces in one: FIT Core, a multi-language function runtime (Java, Python, C++) with hot-swappable plugins; WaterFlow, an engine for composing streaming business logic; and FEL, primitives that wrap models, vector stores, and tool calls so RAG pipelines and agents can live inside a Java codebase instead of a separate service. The docs show retrieve and agent examples that read like ordinary Java fluent APIs, which makes it easy for a team to stand up a retriever or an agent loop without learning a new language. It also handles the deployment plumbing, routing calls locally or over RPC depending on whether you run it as a monolith or distributed services. For shops on Java or Spring that want LLM features embedded rather than bolted on as a Python sidecar, it is worth comparing against whatever you already use.