This project provides an innovative AI-enhanced solution that not only automates code conversion between programming languages but also generates comprehensive test cases to validate the converted code. By leveraging Amazon Nova Premier through Amazon Bedrock, this solution delivers a complete 4-phase AI workflow that transforms code conversion from a risky manual process into a reliable, automated pipeline with built-in quality assurance.
The solution implements a comprehensive AI workflow:
Phase 1: AI-Powered Analysis
- Intelligent complexity assessment of source code
- Pattern identification and dependency analysis
- Conversion challenge prediction
- Context understanding for optimal conversion strategy
Phase 2: AI-Enhanced Conversion
- Context-aware code translation using analysis insights
- Intelligent chunking based on code complexity
- Structure and logic preservation
- Target language idiom optimization
Phase 3: AI-Powered Validation
- Quality assessment of converted code
- Confidence scoring (0-100 scale)
- Issue identification and reporting
- Conversion accuracy verification
Phase 4: AI Test Generation
- Comprehensive test case creation
- Framework-specific test generation (pytest, junit, jest, etc.)
- Edge case and integration test coverage
- Dual persistence (local + S3 storage)
- Execution instruction generation
Traditional code conversion approaches face a critical gap: even if code is successfully converted, how do you verify that it works correctly? This solution addresses that challenge by providing:
- Complete End-to-End Automation - From initial analysis to final test case generation
- AI-Powered Quality Assurance - Built-in validation with confidence scoring
- Production Readiness - Generated tests provide validation needed for confident deployment
- Reduced Time-to-Deployment - Significantly reduces time from conversion to production
- AI-Enhanced Code Conversion - Context-aware conversion using Amazon Nova Premier
- Automated Test Generation - Comprehensive, executable test cases for validation
- Multi-Language Support - Python, Java, JavaScript, C++, C#, Go, Scala, PHP, R, Bash, PowerShell, SQL, HTML, CSS, TypeScript, Objective-C
- Intelligent Processing - AI analysis drives optimal conversion strategies
- Dual Test Persistence - Tests saved locally (
generated_tests/directory) and uploaded to S3 - Framework-Specific Tests - pytest, junit, jest, NUnit, MSTest compatible test generation
- Real-time Progress Tracking - Monitor all 4 phases with detailed metrics
- Confidence Scoring - AI-powered quality assessment (0-100 scale)
- Cost Estimation - Token usage and cost tracking across all phases
- Error Handling and Logging - Comprehensive logging with Amazon CloudWatch integration
- AWS Account with appropriate permissions to access S3 and Amazon Bedrock
- Python 3.9+ installed
- AWS CLI configured with appropriate credentials
Your AWS credentials need the following permissions:
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Action": [
"s3:PutObject",
"s3:GetObject",
"s3:ListBucket",
"s3:ListAllMyBuckets",
"bedrock:InvokeModel",
"logs:CreateLogGroup",
"logs:CreateLogStream",
"logs:PutLogEvents"
],
"Resource": [
"arn:aws:s3:::*",
"arn:aws:bedrock:*:*:foundation-model/*",
"arn:aws:logs:*:*:*"
]
}
]
}The following packages are included in requirements.txt:
- boto3
- streamlit
- watchtower
- botocore
-
Clone the repository:
git clone https://github.com/aws-samples/code-conversion-using-gen-ai.git cd code-conversion-using-gen-ai -
Install required packages:
pip install -r requirements.txt
-
Configure AWS credentials: Ensure your AWS credentials are properly configured with access to S3 and Amazon Bedrock.
Start the Streamlit application with the complete 4-phase AI workflow:
streamlit run app.py-
S3 Configuration:
- Select source and target S3 buckets
- Optionally specify a folder prefix to limit scope
-
AI Conversion Template:
- Choose from pre-configured AI templates for optimal conversion
- Or select "Custom" for specialized conversion patterns
-
Language Selection:
- Select source and target programming languages
- Templates automatically configure optimal language pairs
-
File Discovery:
- Click "Scan S3 Bucket for Files" to discover source files
- Review the list of files to be processed
-
Start AI Workflow:
- Click "Start Gen AI Workflow" to begin the 4-phase process
- Monitor progress through Analysis → Conversion → Validation → Test Generation
- View real-time metrics including confidence scores and token usage
-
Review Results:
- Examine the comprehensive workflow summary
- Access generated test files locally and in S3
- Review AI confidence scores and validation results
The solution generates comprehensive test suites with the following capabilities:
- Unit Tests - Individual function and method testing
- Integration Tests - Component interaction validation
- Edge Cases - Boundary condition testing
- Error Handling - Exception and error scenario testing
- Python: pytest-compatible test files with fixtures and assertions
- Java: JUnit test classes with proper annotations
- JavaScript: Jest test suites with mocking capabilities
- C#: NUnit or MSTest compatible test classes
- Local Storage: Tests saved in
generated_tests/directory - S3 Integration: Automatic upload to target bucket
- Organized Structure: Tests organized by source file and timestamp
You can configure the following settings in app.py:
- Model Configuration: Amazon Nova Premier model settings
- Logging: CloudWatch Log Group and Stream names
- Processing: Chunk sizes and parallel processing options
- Test Generation: Framework preferences and test types
Converting 1,083 lines of Java code to Python with comprehensive test generation:
- Manual Process: 40-60 hours (including test case creation)
- AI-Enhanced Solution: Under 8 minutes with complete test coverage
- Processing Rate: ~2,500 tokens per minute across all 4 phases
The solution provides comprehensive monitoring through:
- Amazon CloudWatch Integration - Detailed logging across all 4 phases
- Real-time Metrics - Progress tracking with token usage and costs
- AI Confidence Scores - Quality assessment for each conversion
- Test Generation Metrics - Success rates and test case counts
- Performance Analytics - Processing times and throughput metrics
The solution runs on AWS compute services (EC2, EKS, ECS) and integrates with:
- Amazon S3 - Source and target file storage
- Amazon Bedrock - AI model access (Nova Premier)
- Amazon CloudWatch - Logging and monitoring
- Streamlit - User interface and workflow management
To avoid ongoing charges, delete the following AWS resources:
- S3 buckets (including generated test files)
- IAM roles and policies
- EC2 instances or other compute resources
- CloudWatch log groups
Looking forward to enhance the solution with future enhancements including:
- CI/CD Integration - Automated test execution in pipelines
- Automated Test Execution - Immediate feedback on conversion quality
- Enterprise Integration - Advanced workflow management and reporting
This project is licensed under the MIT License. See the LICENSE file for details.
- Amazon Bedrock - For providing Amazon Nova Premier AI capabilities
- Streamlit - For the excellent Python web application framework
- AWS Services - For scalable, reliable cloud infrastructure