opensolon/solon-ai

Java AI application development framework (supports LLM-tool,skill; RAG; MCP; Agent-ReAct,Team-Agent). Compatible with java8 ~ java25. It can also be embedded in SpringBoot, jFinal, Vert.x, Quarkus, and other frameworks.

What it solves

Solon AI is a Java-native development framework designed to help Java developers build production-grade AI applications. It provides a unified way to integrate Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), and autonomous agents into Java ecosystems like SpringBoot, Vert.X, and Quarkus, filling a gap for those who prefer Java over Python-based frameworks like LangChain.

How it works

The framework operates through several core modules:

  • ChatModel: A unified interface that uses "dialects" to adapt to different LLM providers (OpenAI, Gemini, Claude, Ollama, DeepSeek, etc.).
  • RAG Pipeline: A full-link system including document loaders, splitters, embedding models, and reranking models to build knowledge bases.
  • MCP Integration: Deep support for the Model Context Protocol (MCP), allowing for cross-platform sharing of tools, resources, and prompts via both server and client implementations.
  • Agent Orchestration: Uses graph-driven flows to implement ReAct (Reasoning and Acting) loops and multi-agent team collaboration based on preset protocols.
  • Ai Flow: A YAML-based orchestration system that allows developers to define AI workflows (similar to low-code platforms like Dify) without writing extensive code.

Who it’s for

Java developers and enterprise software engineers who want to build AI-powered agents, intelligent assistants, or RAG-based knowledge bases within a Java environment (supporting JDK 8 through 25).

Highlights

  • Java Ecosystem Integration: Seamlessly works with Solon, SpringBoot, Vert.X, and Quarkus.
  • Multi-Model Support: Unified API for diverse LLM providers via dialect adaptation.
  • Graph-Driven Agents: Transforms complex agent reasoning into observable and governable computation flow graphs.
  • MCP Support: Native integration with the Model Context Protocol for tool and resource sharing.
  • Low-Code YAML Flows: Ability to arrange RAG and model calls via YAML configuration.