Agentic SDLC

We designed four agents to automate the SDLC. These were: Requirements Agent: Understand requirements from requirements doc Design Agent: Create high level design doc Software Development Agent: Generate codebase to build PoC (small project) Code Test Agent: Generate code tests The workflow steps we followed were as follows: a. Requirements Gathering Task: Extract key requirements from a document. Goal: Create a concise summary of the CRM system's required features. Outcome: Defines the project scope (authentication, CRUD operations, task management, reporting). b. High-Level System Design Task: Design the architecture of the CRM system. Diagrams Generated: Use Case Diagram Class Diagram Entity-Relationship Diagram (ERD) UI Design for Dashboard Outcome: A document detailing the architecture, components, and visual diagrams of the system. c. Code Generation Task: Develop the Code for the system. Goal: Create functional code that implements core features. Outcome: Working code implementing authentication, database operations, and reporting. d. Code Testing Task: Run test cases to verify code functionality. Goal: Ensure the system meets the requirements and works as expected. Outcome: A detailed test report highlighting results and potential issues. Future Work could include: Improvements in Design Diagrams: Explore more AI-driven tools for automated generation of detailed design diagrams. Customization: Enable more advanced configurations for tasks such as adding new agents or expanding CRM functionality. Deployment: Plan for deployment of the final CRM system in a production environment. We also considered building an agentic workflow for MLOps, but ultimately decided on the SDLC.