Emend

EMEND - PROJECT OVERVIEW

INSPIRATION

The idea behind Emend came from a common problem in software development: fixing one bug can sometimes create another bug somewhere else in the project.

Most AI coding tools focus on fixing the reported issue but may not check whether the changes affect other parts of the code.

We wanted to build a smarter tool that not only fixes bugs but also checks whether the fixes are safe.

This led us to create Emend, an AI-powered automatic code repair system.

WHAT IT DOES

Emend automatically finds bugs, understands their causes, generates fixes, and tests those fixes to make sure they don't create new problems.

Main features:

Automatic Bug Detection: Finds and reproduces errors in the code. Impact Analysis: Identifies which other parts of the project might be affected. AI-Powered Bug Fixing: Uses NVIDIA Nemotron to understand problems and generate fixes. Multiple Testing Levels: Tests fixes and checks related parts of the project. Automatic Recovery: If a fix creates another bug, Emend can undo it and try another solution. Risk Reporting: Identifies risky code that does not have enough test coverage. HOW WE BUILT IT

We used the following technologies:

Backend: Python and FastAPI. AI Integration: NVIDIA Nemotron through Nebius Token Factory. Impact Analysis: AST parsing and NetworkX. Testing System: Python subprocesses and Docker. Frontend: React, TypeScript, Vite, and Tailwind CSS. Live Updates: Server-Sent Events (SSE).

We also added security features to prevent unsafe code execution, secret leaks, and other security problems.

CHALLENGES WE FACED

Our biggest challenge was making sure that fixing one bug did not create another bug somewhere else.

We solved this by creating an impact analysis system that checks related parts of the code.

Another challenge was trusting AI-generated fixes. We built a testing system to check whether the suggested fixes actually work.

We also created an automatic recovery system that can undo failed fixes and try alternative solutions.

Balancing speed, AI costs, security, and reliability was another challenge.

ACCOMPLISHMENTS WE ARE PROUD OF Built an AI-powered automatic bug-fixing system. Created an impact analysis engine. Developed a multi-level testing system. Implemented automatic rollback for failed fixes. Built an interactive dashboard. Combined AI with automated testing to improve reliability.

Our biggest achievement is that Emend doesn't just fix bugs; it also checks whether the fixes are safe.

WHAT WE LEARNED

While building Emend, we learned:

How to integrate NVIDIA Nemotron. How to analyze code using ASTs and dependency graphs. How to build AI agents for automatic bug fixing. Why testing is important for AI-generated code. How to execute code safely. How to build real-time dashboards using React and FastAPI. How to balance performance, security, and AI costs.

We learned that a good AI tool needs more than an intelligent model. It also needs proper testing and security.

WHAT'S NEXT FOR EMEND?

Our future plans include:

More Programming Languages: Supporting additional programming languages. GitHub Integration: Connecting Emend with GitHub Actions and CI/CD pipelines. Better Code Understanding: Improving support for large projects. Smarter AI Agents: Improving bug detection and code generation. Team Collaboration: Adding shared debugging history and team features. Better Performance: Making bug fixing faster and reducing AI costs.

Our long-term goal is to make Emend a trusted AI debugging assistant that helps developers fix bugs faster without affecting software quality.