Inspiration
Both of us have faced the same problem at our day jobs - debugging production issues takes a lot of developer's time which is best spent on new feature work (not on fixing bugs). It's also business critical - any application downtime may directly affect the company's clients in a negative way, potentially resulting in churn and money being lost. We wanted to see if we can cut down this downtime and fixing time, so that the developers can focus on what's important and not what's urgent.
What it does
It suggests fixes by creating Github pull requests based on issues that appear in Sentry - a popular error reporting tool for tech businesses. Our AI agent is integrated with Sentry and Github for a seamless and fast user experience.
How we built it
We leveraged three APIs:
- Sentry - for incident detection and fetching error details
- OpenAI - for suggesting fixes, provided by GPT-4
- Github - for automating fixes by creating pull requests
Challenges we ran into
Sentry API is not very mature and we ran into some unexpected, non-deterministic issues with the API but we managed to resolve those through trial and error.
Accomplishments that we're proud of
Creating an autonomous agent that is actually useful and solves a real-world problem faced by a vast majority of professional developers every day.
What we learned
- how to work under a very tight deadline
- how to communicate clearly
- how to focus on what's important
- how good LLMs are at fixing issues by reasoning
What's next for PRomptFix
- integrate with more error reporting tools and services
- figure out a go-to-market strategy
- polish the product for a flawless user experience
Built With
- cloudflare
- django
- flask
- github
- gpt-4
- html
- javascript
- openai
- postgresql
- python
- render
- ruby
- ruby-on-rails
- sentry
- tailwind
Log in or sign up for Devpost to join the conversation.