PACGBI-Tool-Dev

PACGBI (Pipeline for Automated COBOL Generation from Backlog Items) is an AI-powered tool that automates debugging and maintenance of COBOL systems by translating GitHub issues into code-level fixes and documentation. By integrating graph theory, and large language models (LLMs), PACGBI locates affected COBOL functions, generates patches, creates UML diagrams, and opens ready-to-merge pull requests—all following a fully automated GitHub-based workflow.

---

Features

• LLM-based Fix Generation - using Mistral API
• Graph-based Code Analysis - for function relevance scoring
• GitHub Issue Integration - using the REST API
• UML Diagram Generation - for legacy code visualization
• Automatic Pull Requests - with commit-ready patches
• GitHub CLI Extension - to easily run the tool from your terminal

---

Repo Structure

Papers - Refernce Paper Tool - The files containing the helper classes used in the pipeline. cobal-test - Contains the test files on which the model was evaluated

Getting Started

Follow these steps to set up and run the project locally:

1. Clone the Repository:

   git clone https://github.com/adithya-ananth/PACGBI-Tool-Dev.git
   cd PACGBI-Tool-Dev

2. Install required dependencies in a new virtual environmnet:

   python -m venv env
   ./env/Scripts/activate
   pip install -r requirements.txt
   

3. Run the pipeline.py file in the Tool Directory to execute the code via the terminal:

   cd Tool
   python pipeline.py
   

Alternatively, install and run as a GitHub extension:

pip install git+https://github.com/adithya-ananth/PACGBI-Tool-Dev.git
pacgbi

If you find this tool helpful in your research or projects, please cite:
PACGBI: A Pipeline for Automated COBOL Generation from Backlog Items.
Adithya Ananth, Anirudh Arrepu, Dhyanam Janardhana, Gadepalli Srirama Surya Ashish, Srikar Vilas Donur, and Sudhanva Bharadwaj BM.

Team Mentor: Dr. Sridhar Chimalakonda, IIT Tirupati