PatchPilot AI

PatchPilot AI is a generic agentic code review, debugging, and improvement copilot for developers working across C++, Java, JavaScript/TypeScript, Python, and C#. Developers often work with both simple scripts and complex systems. Bugs can come from logic mistakes, unsafe memory operations, async issues, security vulnerabilities, performance bottlenecks, or maintainability problems. PatchPilot AI helps by turning uploaded code into a structured engineering report. The system uses a multi-agent workflow: - Language & Structure Agent: detects languages, file structure, and review scope. - Bug & Correctness Agent: identifies risky code patterns, edge cases, unsafe memory usage, async/lifetime issues, and likely logic problems. - Security Agent: flags injection risks, unsafe eval/exec usage, DOM XSS, shell command execution, and hardcoded secrets. - Performance & Complexity Agent: highlights inefficient loops, async misuse, hot-path logging, thread creation, and scalability risks. - Refactor Agent: recommends maintainability improvements and cleaner structure. - Test Planner Agent: generates unit, regression, security, fuzz, load, and integration test ideas. - Executive Summary Agent: prioritizes the most important issues and creates a fix roadmap. The MVP works without API keys using deterministic static-pattern analysis, making the live demo reliable for judges. It also supports optional OpenAI-compatible model mode, so the project can connect to GPT-style coding/reasoning models or to an AMD Developer Cloud endpoint running a model through ROCm/vLLM. PatchPilot AI targets a real developer need: quickly turning code into actionable bug reports, security checks, performance notes, refactor recommendations, and tests.