Replication Artifact for ASE 2026 Submission:CodeAgent

Read full text Summarise with VIRAT
Authors: Anonymous, Anonymous
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

# CodeAgent - ASE Paper Code Repository This is a CodeAgent project for ASE paper, containing code implementations for 6 tasks. ## Project Structure ```CodeAgent/├── README.md # Project documentation├── requirements.txt # Python dependencies├── LICENSE # License file├── .gitignore # Git ignore file├── scripts/ # Main scripts directory│ ├── task1-4/ # Task 1-4 related scripts│ │ └── run_task.ps1 # Unified task execution script│ ├── task5/ # Task 5 related scripts│ │ ├── download_repos.py # Repository download script│ │ └── send_script.py # Send script│ └── task6/ # Task 6 related scripts│ └── auto_task6.py # Auto task 6 script├── data/ # Data files directory│ ├── task1/ # Task 1 data│ │ └── task1.json # Task 1 project data│ ├── task2/ # Task 2 data│ │ └── task2.json # Task 2 project data│ ├── task3/ # Task 3 data│ │ └── task3.json # Task 3 project data│ ├── task4/ # Task 4 data│ │ └── task4.json # Task 4 project data│ ├── task5/ # Task 5 data│ │ ├── task5.json # Task 5 project data│ │ └── projects/ # Task 5 project repositories (download required)│ └── task6/ # Task 6 data│ ├── groundtruth/ # Ground truth UML diagrams│ │ ├── level1/ # Level 1 projects│ │ ├── level2/ # Level 2 projects│ │ ├── level3/ # Level 3 projects│ │ ├── level4/ # Level 4 projects│ │ └── level5/ # Level 5 projects│ └── projects/ # Task 6 project repositories│ ├── level1/ # Level 1 projects│ ├── level2/ # Level 2 projects│ ├── level3/ # Level 3 projects│ ├── level4/ # Level 4 projects│ └── level5/ # Level 5 projects├── prompt/ # Prompt templates directory│ ├── task1/ # Task 1 prompts│ │ ├── task1_generate.txt # Generation prompt│ │ └── task1_inference.txt # Inference prompt│ ├── task2/ # Task 2 prompts│ │ ├── task2_generate.txt # Generation prompt│ │ └── task2_inference.txt # Inference prompt│ ├── task3/ # Task 3 prompts│ │ ├── task3_generate.txt # Generation prompt│ │ └── task3_inference.txt # Inference prompt│ ├── task4/ # Task 4 prompts│ │ ├── task4_generate.txt # Generation prompt│ │ └── task4_inference.txt # Inference prompt│ ├── task5/ # Task 5 prompts│ │ └── task5.txt # Task 5 prompt template│ └── task6/ # Task 6 prompts│ └── task6.txt # Task 6 prompt template├── output/ # Output results directory (created at runtime)│ ├── task1/ # Task 1 output│ ├── task2/ # Task 2 output│ ├── task3/ # Task 3 output│ ├── task4/ # Task 4 output│ ├── task5/ # Task 5 output│ └── task6/ # Task 6 output└── evaluation/ # Evaluation scripts directory ├── task1-5/ # Task 1-5 evaluation │ └── eva.py # Evaluation script ├── task6/ # Task 6 evaluation │ └── uml_evaluator.py # UML evaluation script └── README.md # Evaluation documentation``` ## Task Description ### Task 1- **Description**: Repository analysis task 1- **Scripts**: PowerShell scripts supporting multiple AI models (agent, claude, codex, elc)- **Data**: JSON format prompt data (no actual repository code included) ### Task 2- **Description**: Repository analysis task 2- **Scripts**: PowerShell scripts supporting multiple AI models (agent, claude, codex, elc)- **Data**: JSON format prompt data (no actual repository code included) ### Task 3- **Description**: Repository analysis task 3- **Scripts**: PowerShell scripts supporting multiple AI models (agent, claude, codex, elc)- **Data**: JSON format prompt data (no actual repository code included) ### Task 4- **Description**: Repository analysis task 4- **Scripts**: PowerShell scripts supporting multiple AI models (agent, claude, codex, elc)- **Data**: JSON format prompt data (no actual repository code included) ### Task 5- **Description**: Bug detection task- **Scripts**: Python scripts for template processing- **Data**: Contains problem descriptions (no actual repository code included) ### Task 6- **Description**: UML class diagram generation task- **Scripts**: Python scripts supporting multi-level repository processing- **Data**: Hierarchically organized repository structure with actual project code ## Installation and Usage ### Quick Installation ```bash# Install dependencies using pippip install -r requirements.txt``` ### Running Tasks #### Task 1-4: Repository Analysis Tasks **Note**: These tasks use prompt-based analysis and do not require downloading actual repository code. ```bashcd scripts/task1-4 # Run with default settings (agent model)./run_task.ps1 # Run specific task with different models./run_task.ps1 -StartTask 1 -Model agent # Task 1 with agent model./run_task.ps1 -StartTask 2 -Model claude # Task 2 with claude model./run_task.ps1 -StartTask 3 -Model codex # Task 3 with codex model./run_task.ps1 -StartTask 4 -Model elc # Task 4 with elc model # Run from specific repository within a task./run_task.ps1 -StartTask 1 -StartRepo "repo_name"``` #### Task 5: Bug Detection Task **Important**: Before running Task 5, you need to download the project repositories. #### Download Project Repositories Task 5 requires downloading actual project repositories based on the information in `task5.json`: ```bashcd scripts/task5 # Download all projects (requires git and internet connection)python download_repos.py # Or download specific projectpython download_repos.py --instance_id "SiSaR-Lab__EduLite-271"``` #### Run Bug Detection ```bashcd scripts/task5 # Run the main scriptpython send_script.py``` #### Task 6: UML Generation Task ```bashcd scripts/task6 # Generate UML diagrams for all levelspython auto_task6.py # Generate for specific level (1-5)python auto_task6.py --level 1``` ### Evaluation #### Tasks 1-5 Evaluation ```bashcd evaluation/task1-5 # Evaluate all taskspython eva.py # Evaluate specific taskpython eva.py --task 1 # Task 1 onlypython eva.py --task 2 # Task 2 onlypython eva.py --task 3 # Task 3 onlypython eva.py --task 4 # Task 4 onlypython eva.py --task 5 # Task 5 only``` #### Task 6 Evaluation (UML Generation) ```bashcd evaluation/task6 # Single file evaluationpython uml_evaluator.py generated_uml.json groundtruth_uml.json # Batch evaluation for all projectspython uml_evaluator.py ../../output/task6/generation ../../data/task6/groundtruth``` ### Configuration Before running the tasks, ensure you have:- API keys for AI models (agent, claude, codex, elc) configured in environment variables- Required permissions for repository access- Sufficient disk space for output files ## Dependencies - Python 3.8+- PowerShell 7+- Related AI model API access permissions ## License This project follows the MIT License.

Fields of Study & Subjects

Software Engineering Research Software Engineering Techniques and Practices Scientific Computing and Data Management

Publication details

Journal / source
Figshare
Publication date
27 March 2026
Language
EN
Document type
Dataset