🛡️ AI Code Reviewer Agent

A professional-grade AI Code Review Agent built in Java. It automates code analysis against proprietary organizational standards using a Retrieval-Augmented Generation (RAG) architecture.

Designed for Enterprise Environments, it features a sliding window engine for processing legacy monolithic files and a model-agnostic design (Google Gemini + Groq).

---

🏗️ Architecture

The application follows a Clean Architecture pattern, decoupling the AI service layer from business logic.

graph TD
    User["Legacy Code File"] -->|Input| Engine["Sliding Window Engine"]
    Standards["CompanyStandards.txt"] -->|Ingest| RAG["Vector Store (Memory)"]
    
    Engine -->|Chunk 1..N| AI["AI Agent"]
    RAG -->|Context Injection| AI
    
    AI -->|Request| LLM["LLM Provider (Groq/Gemini)"]
    LLM -->|JSON Review| AI
    AI -->|Report| Output["Final Analysis"]

Loading

---

🚀 Key Features

1. 🧠 RAG-Powered Compliance: Ingests a custom CompanyStandards.txt rulebook to enforce specific internal guidelines (e.g., "Always use Loggers," "Snake_case variables") rather than generic advice.
2. 🪟 Sliding Window Engine: Uses a custom Deque-based algorithm to process large files (>10,000 lines) by segmenting context into overlapping chunks, preventing LLM context window overflow.
3. ⚡ Hybrid AI Model: Optimizes cost and speed by using Google Gemini for Vector Embeddings and Groq (Llama 3.3) for high-speed inference.

---

🛠️ Project Structure

src/main
├── java/org/example/reviewer
│   ├── AgentFactory.java        # Wires RAG, Embeddings, and Chat Models
│   ├── App.java                 # Main Entry Point
│   ├── CodeAnalysis.java        # Java Record (Data Model)
│   └── SlidingWindowEngine.java # Algorithmic Logic for large files
└── resources
    ├── CompanyStandards.txt     # Your custom coding rules
    └── BadCode.java             # Test file for analysis

---

⚙️ Step-by-Step Setup

Step 1: Prerequisites

• Java 21 (JDK)
• Maven 3.x
• API Keys:
  • Google AI Studio (Free) - For Embeddings.
  • Groq Console (Free) - For Llama 3 Inference.

Step 2: Dependencies (pom.xml)

Add the following LangChain4j dependencies to your pom.xml file:

<dependencies>
    <dependency>
        <groupId>dev.langchain4j</groupId>
        <artifactId>langchain4j</artifactId>
        <version>0.35.0</version>
    </dependency>

    <dependency>
        <groupId>dev.langchain4j</groupId>
        <artifactId>langchain4j-google-ai-gemini</artifactId>
        <version>0.35.0</version>
    </dependency>

    <dependency>
        <groupId>dev.langchain4j</groupId>
        <artifactId>langchain4j-open-ai</artifactId>
        <version>0.35.0</version>
    </dependency>

    <dependency>
        <groupId>org.slf4j</groupId>
        <artifactId>slf4j-simple</artifactId>
        <version>2.0.7</version>
    </dependency>
</dependencies>

Step 3: Configure Environment

Set your API keys as environment variables (or configure them in App.java):

• GOOGLE_API_KEY
• GROQ_API_KEY

Step 4: Define Rules

Create src/main/resources/CompanyStandards.txt and add your rules:

1. SECURITY: Never use 'System.out.println'. Use a Logger.
2. STYLE: Variables must be in snake_case.
3. PERFORMANCE: Use StringBuilder for string concatenation.

---

🏃 Usage

1. Compile the Project

mvn clean install

2. Run the Application

You can run the App class directly from IntelliJ/Eclipse, or use Maven:

mvn exec:java -Dexec.mainClass="org.example.reviewer.App"

3. Sample Output

📚 [RAG] Found standards at: /path/to/resources/CompanyStandards.txt
📚 [RAG] Ingesting Company Standards...
🤖 [AI] Connecting to Groq (Llama 3.3)...
📂 Analyzing resource: /path/to/resources/BadCode.java
🚀 [Engine] Starting analysis...

   Processing lines 0-20... ✅ Score: 40
   Processing lines 15-35... ✅ Score: 65

==========================================
       FINAL REPORT: BadCode.java
==========================================
Average Quality Score: 52/10
Total Policy Violations: 3
------------------------------------------
🔴 Violation: Used 'System.out.println' (Security Risk)
🔴 Violation: Variable 'myVar' is camelCase (Style Rule)
==========================================

---

Built with ❤️ using LangChain4j