FIT Framework is a Java-based AI development framework providing multi-language function engine, flow orchestration, and a Java-oriented LangChain alternative, with native/Spring dual-mode execution and plugin hot-plug deployment.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
FIT Framework is a Java enterprise AI development framework that offers a multi-language function engine (FIT), a streaming orchestration engine (WaterFlow), and a Java ecosystem LangChain substitute (FEL). It supports native and Spring dual-mode operation, plugin hot-plug deployment, and unified deployment of large models with business systems.
How it works
The core comprises three parts:
- FIT Core: language-agnostic function computation with plug-in hot-swapping and intelligent sharding/ deployment that routes calls locally or via RPC.
- WaterFlow Engine: visual orchestration with declarative APIs, enabling modular composition of microflows and cross-system long transactions.
- FEL (FIT Expression for LLM): Java-based primitives for models, knowledge bases, and toolchains, to integrate AI capabilities into Java tech stacks.
The design emphasizes
- convention over configuration to reduce boilerplate, automatic deployment relationships, protocol transparency, and self-managing resources via plugin dependency injection and built-in service discovery and circuit breaking.
Getting started
- Build commands:
# 仅构建 Java 框架
mvn clean install
# 全局构建（Java 框架 + Sandbox CLI），默认跳过测试
./build.sh
# 全局构建（含测试）
./build.sh --with-test
- Output directory structure includes a fit startup command at build/bin/fit and a sandbox at build/bin/sandbox, with corresponding jars and libs.
- Start command:
./build/bin/fit start
-
Environment setup notes describe configuring a FIT core framework directory and an environment variable FIT to point to it, plus creating a plugin directory to house plugins and start services via fit start.
-
Quick starts link to:
- FIT quick-start guide
- WaterFlow quick-start guide
- FEL quick-start guide
Recent releases
Latest releases show updates to the FIT Function Platform with enhancements and bug fixes:
- v3.6.4 (2026-03-13): enhancements include upgrading central-publishing-maven-plugin to 0.10.0 and a docker change renaming noAuthHint to setupHint to avoid CodeQL warnings; bugfix entries exist but truncated in the listing.
- v3.6.3 (2026-02-24): enhancements include refactoring sandbox scripts to TypeScript CLI entry point, introducing a declarative AI tool registry to reduce per-tool duplication, and sandbox support for Gemini CLI/OpenCode/Codex.
- v3.6.2 (2026-01-09): enhancements include upgrading swagger-ui to 5.31.0 and a couple of framework-related changes.
- v3.6.1 (2025-12-05): enhancements include fixes to Python process restart mechanism and scheduler-related AttributeError, plus docker-related updates.
- v3.6.0 (2025-11-13): overview of improvements across FIT Function Platform, FEL, and WaterFlow scheduling.
Traction
The repository has 2113 stars and 334 forks, with 55 open issues, and was created in 2025-02-22. It is primarily Java language and MIT license. Last push was 2026-03-13.
Caveats
License: MIT. Language: Java. Created: 2025-02-22. Last push: 2026-03-13. Open issues: 55. Stars: 2113. Forks: 334.
Behind the repo
Not provided in the README excerpt.