Teaching Java devs to build ReAct agents, one date plan at a time

What it does

This is the codebase for a Chinese tutorial series that teaches Java developers to build AI applications using Spring Boot 3, Java 21, and Spring AI. It centers on two applications: an “AI Love Master” chatbot that offers romantic advice through multi-turn conversation, RAG retrieval, and tool calling; and YuManus, a ReAct-pattern agent that autonomously searches the web, manipulates files, and generates PDF reports to complete tasks like planning a date. The curriculum runs from model integration to Serverless deployment.

The interesting bit

The author uses a dating-advice chatbot as a vehicle for serious engineering lessons. By wrapping dry enterprise patterns—vector search, MCP protocol wiring, and structured output parsing—inside a whimsical domain, the project manages to teach Spring AI without feeling like another generic CRM tutorial.

Key highlights

• Implements RAG with PGVector, covering document chunking, embedding storage, retrieval strategies, and hallucination mitigation.
• Covers Tool Calling and MCP service development, equipping the agent with web search, file operations, terminal access, and PDF generation.
• Builds a ReAct-based autonomous agent (YuManus) that plans and executes multi-step tasks.
• Taught as part of a paid bootcamp that bundles video lectures, resume guidance, and an AI-themed interview question bank.
• Uses Spring AI as the primary framework, with LangChain4j and Ollama local deployment in the mix.

Caveats

• The repository is explicitly a teaching project, and the README is largely a landing page for a paid course platform with limited standalone technical documentation.
• The full narrative—architecture decisions, bug fixes, and advanced tuning—is delivered through video, so the code alone may read like lecture slides without the lecture.

Verdict

A solid reference if you are a Java backend developer trying to map Spring Boot concepts onto AI engineering. Look elsewhere if you need a drop-in agent framework or a fully open-source, self-documented product.