Posts on JavaDev AI - Java + AI in Practice

How to Build an AI Code Review Agent with LangChain4j and Spring Boot

Sun, 28 Jun 2026 00:00:00 +0000

Automated code review is one of the most practical applications of AI in software development. In this guide, you’ll learn how to build an AI-powered code review agent using LangChain4j and Spring Boot that integrates with GitHub to automatically review pull requests.

Why Build an AI Code Review Agent?

Manual code reviews are time-consuming and inconsistent. Studies show that developers spend up to 6 hours per week on code reviews. An AI agent can:

LangChain4j vs Spring AI: Which Framework Should Java Developers Choose in 2026?

Sat, 27 Jun 2026 00:00:00 +0000

Java developers building AI applications face a critical choice: LangChain4j or Spring AI? Both frameworks enable LLM integration, but they take fundamentally different approaches. After building production applications with both, here’s an honest comparison to help you decide.

Quick Answer

Choose LangChain4j if you want maximum flexibility, mature Agent/RAG support, and don’t want to be locked into the Spring ecosystem.
Choose Spring AI if you’re already deep in the Spring ecosystem and want tight integration with Spring Boot auto-configuration.

For most new AI projects in 2026, LangChain4j is the safer bet. Here’s why.

Spring Boot AI Integration: Complete Guide to Adding LLM Capabilities to Your Application

Fri, 26 Jun 2026 00:00:00 +0000

Adding AI capabilities to your Spring Boot application doesn’t require rebuilding from scratch. In this comprehensive guide, you’ll learn how to enhance an existing Spring Boot application with LLM-powered features using LangChain4j.

What You’ll Build

A customer support assistant that can:

Answer questions about your product using documentation (RAG)
Process natural language commands via tool calling
Maintain conversation context across requests

Step 1: Add Dependencies

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15


<dependencies>
 <!-- LangChain4j with OpenAI -->
 <dependency>
 <groupId>dev.langchain4j</groupId>
 <artifactId>langchain4j-open-ai-spring-boot-starter</artifactId>
 <version>0.36.2</version>
 </dependency>
 
 <!-- For RAG with PgVector -->
 <dependency>
 <groupId>dev.langchain4j</groupId>
 <artifactId>langchain4j-pgvector</artifactId>
 <version>0.36.2</version>
 </dependency>
</dependencies>

Step 2: Configure LLM Connection

 1
 2
 3
 4
 5
 6
 7
 8
 9
10


# application.yml
langchain4j:
 open-ai:
 chat-model:
 model-name: gpt-4o-mini
 temperature: 0.3
 max-tokens: 2000
 timeout: 30s
 log-requests: true
 log-responses: true

For cost optimization, use gpt-4o-mini for most tasks and gpt-4o only for complex reasoning.

10 MyBatis SQL Injection Vulnerabilities AI Can Catch That Humans Miss

Thu, 25 Jun 2026 00:00:00 +0000

MyBatis is the most popular ORM framework in the Java ecosystem, powering millions of applications. But its flexibility comes with a dangerous gotcha: ${} vs #{} syntax. One is safe, the other is not — and the difference is a single character.

Here are 10 MyBatis SQL injection patterns that slip past human reviewers but an AI code review agent catches instantly.

The Core Problem: `${}` vs `#{}`

1
2
3
4
5
6
7
8
9


<!-- SAFE: #{} uses PreparedStatement parameter binding -->
<select id="findById" resultType="User">
 SELECT * FROM users WHERE id = #{id}
</select>

<!-- VULNERABLE: ${} directly interpolates the string -->
<select id="findById" resultType="User">
 SELECT * FROM users WHERE id = ${id}
</select>

The difference: #{id} generates WHERE id = ? with parameter binding, while ${id} generates WHERE id = 1 OR 1=1 with direct string interpolation.

Building a Profitable AI SaaS as a Solo Developer: Lessons from $5k MRR

Wed, 24 Jun 2026 00:00:00 +0000

After 8 months of building an AI code review tool as a solo developer, I hit $5,000 MRR (Monthly Recurring Revenue). This article shares the strategy, technical decisions, and lessons learned along the way.

The Starting Point

I’m a 10-year Java developer. I knew AI was transforming software development, and I wanted to build something useful — not a chatbot wrapper, but a tool that solves a real pain point.