RAG Platform (Spring AI Multi-Module)

Default language: Chinese. Click the button above to switch.

默认语言：中文。点击上方按钮切换语言。

Why This Project Matters

gengzi/rag is an Apache-2.0 open-source, Spring AI based multi-module RAG and Agent platform for the Java ecosystem.

Many RAG and Agent projects are Python-first or demo-oriented. This repository focuses on enterprise-style Java / Spring backend engineering and provides a modular reference implementation for knowledge ingestion, vector retrieval, SSE streaming chat, message persistence, Agent orchestration, GraphRAG, MCP memory/tool integration, and multi-agent collaboration.

The goal is to help Java backend teams build production-oriented AI applications without starting from scattered demos.

What This Repo Contains

• Multi-module Spring AI RAG platform
• Knowledge ingestion + vector retrieval + chat memory
• Agent orchestration (DeepResearch / PPT / Excalidraw / Text2SQL)
• Graph-enhanced retrieval (Neo4j)
• MCP server/client integration for tool and memory extension
• Team-style multi-agent collaboration sample

Read In Order

1. Full Chinese architecture guide: README.zh-CN.md
2. English quick architecture guide: README.en.md
3. Architecture overview with diagrams: docs/ARCHITECTURE_OVERVIEW.md
4. Roadmap: ROADMAP.md
5. Contribution guide: CONTRIBUTING.md
6. Agent Teams design deep dive: rag-agent-teams/docs/DESIGN.md
7. Graph subsystem design deep dive: rag-graph/docs/SYSTEM_DESIGN.md

Module Ports

• rag-core: 8883
• rag-chat: 8086
• rag-agent: 8889
• rag-graph: 8199
• rag-mcp: 8890
• rag-mcp-client: 8891
• rag-agent-teams: 8080