Sarah Ait Yahia@Sarah_Ait

Problem & Solution Statement Software developers often face challenges when working with unfamiliar or large codebases. Understanding project structure, identifying entry points, tracing dependencies, and locating relevant functions can be time-consuming. Manual code reviews also require significant effort to identify potential security vulnerabilities and code quality issues, making software maintenance more difficult. IBM Developer Copilot Platform is a multi-agent developer assistant designed to simplify codebase understanding and code review through a unified interactive workspace. The platform provides two main modes: Type 1 – Codebase Understanding: Developers upload a ZIP archive of their project. Agent 1 extracts, parses, chunks, and indexes the source code using a Retrieval-Augmented Generation (RAG) approach. Agent 2 analyzes the indexed project to summarize its architecture, technologies, entry points, and module structure. Agent 3 enables developers to ask questions about their codebase and retrieve contextually relevant information. Type 2 – Code Review: Agent 4 performs security and code quality analysis to identify potentially unsafe patterns, such as hardcoded credentials and dangerous code execution. The platform also provides a refactoring workflow through Agent 5, with a testing workflow planned for further development. The target users are software developers, students, development teams, and maintainers who need to understand, review, and maintain existing software projects more efficiently. Users interact with the platform through a Streamlit interface connected to a FastAPI backend. They can upload projects, explore architecture summaries, ask questions, submit code for review, and access the refactoring and testing workflow. What makes our solution distinctive is its combination of specialized agents for codebase exploration and code review in one workspace. Instead of manually navigating files and switching between separate tools, developers can follow a structured workflow from understanding source code to identifying potential issues and exploring improvement opportunities. The current prototype demonstrates codebase indexing, architecture analysis, code-related question answering, and rule-based security review. Refactoring and test generation are currently demonstrated in simulation mode when a real LLM integration is unavailable. Future work includes integrating a real LLM, implementing functional automated refactoring and test generation, and expanding the agents' capabilities.

💻 Online Hackathon | ⏳ Build: 48 hours | 📅 September 25–27, 2026 The official IBM Bob 2.0 hackathon. Build what's next in AI-assisted development, alongside the developers and experts shaping it. ⭐ Get help from experts throughout the event 🤝 Join a team or build solo 💰 Prize pool: $12,000 🧑🏻‍💻 Register before the Kick-Off Stream to start building with Bob