Crash Eye

AI-driven crash detection system: Built to pace the development cycle

Comment
  • Inspiration: Prior to our application's existence, the occurrence of a software crash often prompted users to either close the tab or postpone reporting the issue. This posed a significant obstacle for developers as it made identifying the cause of the crash quite challenging. To address this, our application was developed with a specific functionality: whenever a software crash occurs, it automatically collects and retains all crash-related data within MongoDB in the form of log files. This data is intelligently categorized into three subsets—fatal, critical, and warning—and stored as subfiles within an established dataset structure.

  • What it does: Our innovative approach empowers developers by providing them direct access to these categorized error logs stored in MongoDB. By having crash data neatly organized based on severity, the process of identifying and resolving issues becomes notably more efficient. Consequently, this streamlined approach significantly enhances the speed and effectiveness of the entire software development cycle.

  • How we built it: The project's backend relied on Python, utilizing a logger to capture Crash Eye files. These files underwent classification into distinct parameters utilizing scoring methods from sklearn. Post-classification, the data seamlessly flowed into MongoDB for storage. Developers accessed this stored data from MongoDB, utilizing it to promptly identify crash logs and efficiently resolve emerging issues. Meanwhile, on the frontend, React.js was the chosen framework, empowering the creation of an intuitive user interface. The deployment aspect was handled with Electron.js. Throughout the project, the aim was to harness the potential of the MERN stack, leveraging its capabilities to achieve the most optimal outcomes in terms of efficiency and performance.

  • Challenges we ran into: Reading the crash log files and dump files to classify it using score-based methods. For front-end initially we planned on using express.js but we couldn't figure out the way we should deploy it in desktop locally so we shifted to electron which made the process easier.

  • Accomplishments that we're proud of: We are proud of building this project in 24 hours, we were happy with the output and were successfully able to detect the exact time system info and why it crashed using dummy programs.

  • What we learned: We learned to use different technologies like electron and learned how to connect python to the backend and also learned about several types of violations to the course policies.

  • What's next for Crash Eye: We have planned to enhance and make this project more secure and use similarity algorithms to form clusters of similar crash data with real timeupdates.

Built With

Share this project:

Updates