DevPilot: AI Repository Doctor

Developers working with unfamiliar, complex, or growing codebases often spend significant time understanding project structure, tracing dependencies, identifying potential issues, and figuring out where to add tests. These tasks become especially difficult when information is scattered across multiple files. DevPilot is a repository intelligence and developer productivity platform that helps developers understand and maintain their codebases from a single workspace. Built on a React, Vite, and Tailwind frontend with a FastAPI and SQLAlchemy backend, DevPilot extends an existing development workspace with a Repository Intelligence module. Users can scan an open project and move through repository overview, architecture analysis, dependency visualization, issue detection, test suggestions, and question answering. The scanner analyzes files through an isolated LocalSandbox boundary, ensuring repository file access remains within the project's own workspace. It detects languages and frameworks using project manifests, extracts Python and JavaScript/TypeScript imports to build a dependency graph, and generates an architecture explanation and SVG visualization based on the parsed code. A deterministic set of rules identifies potential code issues. The test suggestion module generates pytest skeletons for supported simple pure functions. Developers can also ask questions about their repository using keyword-based retrieval, with answers grounded in extracted project content and file/line citations. An optional OpenAI API integration can polish architecture explanations and answers, while heuristic and retrieval-based functionality remains available without an API key. The target users include students, independent developers, and software engineering teams who need to understand unfamiliar repositories, investigate potential problems, and identify useful tests. The workflow is designed to reduce repetitive manual code exploration and bring repository understanding into the development environment. What makes DevPilot distinctive is its combination of repository-grounded analysis, dependency visualization, deterministic checks, and cited question answering within an existing isolated workspace, rather than relying on generic answers disconnected from the actual project. The prototype uses React, TypeScript, FastAPI, SQLAlchemy, PostgreSQL, and a dependency-free SVG graph renderer. Its current limitations include rule-based issue detection rather than a full linter, no execution or verification of generated tests, no repository cloning or Git import, and an in-memory scan cache. These limitations are documented in the project documentation.