AI-Powered code reverse

Astah Professional and Astah UML ver.11 enhance existing code reverse capabilities with AI.
AI-Powered Code Reverse adds automatic Sequence Diagram generation and expanded programming language support for code-based diagram generation.

This makes it easier to understand system behavior and structure from code, especially when working with unfamiliar codebases.

Astah AI Code Reverse

Key Features

1. Class Diagram Generation

Automatically generates UML Class Diagrams from source code in languages such as Java, C#, C++, C, Python, Go, TypeScript, and more. (→ Jump to Class Diagram Generation)

2. Sequence Diagram Generation

Automatically generates UML Sequence Diagrams based on the actual execution flow of the code to support deeper analysis. (→ Jump to Sequence Diagram Generation)

Supported Programming Languages

The following languages are supported:

  • Java: Full support for classes, interfaces, enums, annotations, generics, etc.
  • C#: Supports .NET Framework and .NET Core / .NET 5+ projects
  • Python: Supports Python 3 constructs such as classes, decorators, and type annotations
  • TypeScript: Supports interfaces, generics, decorators, and TypeScript-specific features
  • Go: Supports structs, interfaces, method receivers, and more
  • C++: Supports classes, templates, namespaces, and C++ language features
  • JavaScript, Kotlin, Swift, PHP, Ruby, Rust, and others

UseCases

Scenario 1: Understanding an Existing System

  • Quickly generate visual architecture diagrams when onboarding to a project
  • Help new engineers grasp the overall flow when reading code is difficult
  • Understand complex branching or exception handling and method call relationships

Scenario 2: Debugging and Issue Analysis

  • Difficult-to-locate bugs that span multiple methods
  • Unclear execution order or call chain
  • Suspected gaps between implementation and design documentation

Scenario 3: Impact Analysis for New Features

  • Understand collaboration between classes required for a new feature
  • Identify where existing methods are referenced across the codebase
  • Visualize execution flow for specific use cases

Configuration

AI Service Settings

Before using the plugin, configure the API settings for your AI service.

Supported AI Models

  • OpenAI – GPT-5, GPT-5.1, GPT-4 series
  • Anthropic Claude – Claude 4 series (Opus, Sonnet, Haiku)
  • Google Gemini – Gemini Pro, Gemini Flash, Gemini 2.5 series
  • Other OpenAI-compatible services – Any custom service using OpenAI-compatible APIs

Setup Steps

  1. Open the Settings Screen
    • In Astah, go to Tools → AI Code Reverse
      AI Code Reverse Menu


    • The AI Code Reverse dialog will appear, click Option
      AI Code Reverse Option

  2. Add or Edit API Configuration
    • Open a drop-down list for Current Configuration – this has a default set of AI services.
      Current Configuration for AI Code Reverse Service
    • Enter the following information then click Add. The Add button does not work unless you enter Name.
      API Options for Code Reverse
      • Configuration Name: Custom name
      • API URL: Endpoint URL of the AI service
      • API Key: API key provided by the AI service
      • Model Name: Name of the AI model (e.g., gpt-5.1, claude-opus-4-1-20250805, gemini-2.5-pro)

  3. Test the Connection
    • Click Test Connection to verify configuration correctness
      Test Connection
  4. Click OK to save changes.

Class Diagram Generation

Steps

  1. Open the Dialog
    • Go to Tools → AI Code Reverse
      AI Code Reverse Menu

    • The AI Code Reverse dialog opens
      Auto Create Class Diagrams with AI

  2. Select Target for Analysis
    • Target Directory: Choose the directory containing source files (multi-language projects are supported)
    • Include Subdirectories: Enable if needed

  3. Choose Diagram Type
    • Under Target Diagram Type, select Class Diagram

  4. Set Generation Options
    • Auto create class diagram: Automatically create class diagrams per package/namespace (recommended OFF for large projects; you can generate diagrams manually afterward)
    • Show full class details: Generate more detailed diagrams with the appearance of classes, attributes, and operations.
    • User Prompt: Enter custom instructions (e.g., “Ignore all non-public members”)

  5. Start Generation
    • Click OK to begin analysis.
    • The AI analyzes the code and generates a UML static model (classes, attributes, operations, inheritance, interfaces, associations, etc.).
    • Ensure to enable auto create class diagrams to generate diagrams automatically.

Reversing Rules

Based on field types, AI Code Reverse will intelligently determine whether to model something as an attribute or an association.

  • Modeled as attributes:
    • Primitive types
    • Strings
    • Enums
    • Collections of value types (e.g., List<string>, Map<string, int>)

  • Modeled as associations:
    • User-defined types
    • Collections of user-defined types
    • Arrays of user-defined types
section divider

Sequence Diagram Generation

Steps to generate Sequence Diagrams

  1. Open the Dialog
    • Go to Tools → AI Code Reverse
      AI Code Reverse Menu

    • The AI Code Reverse dialog opens
      Auto Sequence Diagram Generation
  2. Select Target for Analysis
    • Target Directory: Choose the directory containing source files (multi-language projects are supported)
    • Include Subdirectories: Enable if needed

  3. Choose Diagram Type
    • Under Target Diagram Type, select Sequence Diagram

  4. Set Sequence Options
    • Entry Methods (optional): e.g., OrderService.processOrder
      • Leave blank to allow automatic detection
      • Learn more about Entry Methods

    • Max Call Hierarchy Depth:
      You can specify the maximum message-call depth for the generated diagram.
      Recommended 3–10

    • User Prompt: Custom analysis rules (e.g., “Ignore all logging calls”)

  5. Start Generation
    • Click OK
    • The plugin analyzes the program flow and generates the Sequence Diagram

Reversing Rules

Specifying a Entry Method

You can specify an “entry method” to let AI know which method to start to reverse.
You can analyze the execution flow of a specific method (e.g., PaymentService.process()).
If you don’t specify the “entry method”, AI automatically detects an appropriate starting point, such as:

  1. Web framework entry points
    • Spring Boot @RestController
    • Flask @app.route
    • Express app.get()
  2. Public API methods
  3. Key business logic methods
  4. Test entry points
    • Unit test or integration test methods

TIPS:
Start by leaving the Entry Methods field blank to allow automatic detection, then refine the analysis by specifying a method.

Multi-Language Call Chain Tracking

The AI can follow cross-language call chains (e.g., Java → Python → Go) and generate multi-language Sequence Diagrams.

Troubleshooting

Issue: Multi-Language Analysis Is Inaccurate

Check:

  • Check if the programming language version is supported
  • Provide clearer user prompts
  • Try a different AI model

Issue: Multi-Language Call Tracking Fails

Check:

  • Ensure all related source code is within the analysis directory.
  • Confirm that network calls use standard protocols (REST, gRPC).
  • Specify the entry method and call path manually.

Issue: Out of Memory in Large Projects

Solution:

  • Increase Astah’s JVM memory allocation.
  • Analyze modules separately instead of the entire project at once.

Usage Tips

  • We recommend starting with smaller projects that use a single programming language.
  • Adjust your prompts based on the generated output and gradually refine them to improve results.

Notes, Cautions, and Safe Use

Handling Confidential Information

  • The handling of information sent to AI services varies depending on the service used and the applicable contract terms. Please review the terms of use, data policies, and settings of each AI service before using it.

Backup Recommendation

  • Before editing models with AI, we strongly recommend creating a full backup of the project.

About AI-Generated Content

  • AI-generated content may not always be complete, correct, or consistent.
  • Always review the output carefully and revise it as needed.
  • Unexpected behavior or issues may occur depending on your prompts, model settings, or service status.

Token Limitations

  • AI services have an upper limit on the number of tokens they can process.
  • Sending or receiving very large specifications or model data in a single request may:
    • Reduce response quality
    • Cause timeout or errors
    • Make processing unstable
  • Try to adjust your prompts so that the exchanged information does not become excessively long.

Updating AI Code Reverse

AI Code Reverse is provided as a plug-in which is already bundled with Astah professional and Astah UML version 11 and later.
When updates with bug fixes or improvements become available, we provide an updated plug-in .jar file. Follow the steps below to update AI Code Reverse to the latest version.

How to Update AI Chat Copilot

  1. Download the latest AI Code Reverse .jar file.
    Currently the latest version is bundled with Astah Professional / Astah UML version 11.

  2. Drag and drop the .jar file into the Astah window while Astah is running.

  3. Restart Astah to apply the update.

Current Version Information

  • Latest Version: 1.0 (Released on 2025/12/17)
  • Bundled Astah Version: Astah Professional / Astah UML Version 11

How to Check Your Installed AI Code Reverse Version

  1. Open [Plugin] → [Installed Plugin].
    Installed Plugins
  2. A list of installed plug-ins will appear.
  3. Locate “Astah AI Code Reverse” and check the Version column.
    AI Code Reverse Plugin List