Inspiration

I wanted to try my best to find an interesting way to do some form of automated observability handling. Given several of the sponsors felt geared towards making developer workflows easier, I decided to try to build a tool to automatically create baseline alerts for a service based on historical metrics and general service implementation analysis.

What it does

Sentralert combines Sentry observability querying and intelligent CodeRabbit reviews to a automatically discover, generate, and apply Sentry alerts.

How we built it

I used Claude Code to do extensive research into each library and discover a simple way to integrate them. I also deployed a simulated FastApi service on Render to be able to more accurately demonstrate Sentralert. I implemented a Sentry Client to handle any Sentry API queries to grab historical metrics and generate alert suggestions. I also used Claude Agents SDK to create an agent equipped with DeepWiki MCP to be able to handle the analysis of the demo service and generate intelligent alerts. I then added a .coderabbit.yaml for more precise CodeRabbit reviews and a GitHub actions script to automatically apply alert suggestions once approved.

Challenges we ran into

Setting up the demo-service and testing the scenarios took a good chunk of time. Unfortunately, that setup isn't totally related to the end product, Sentralert, but was important to me to setup in order to have a more accurate demo.

Designing a simple flow of metrics analysis was also a bit complex. I wasn't sure which aspects or features to delegate to an agent implementation and which to simply codify. In the end, I settled on having Claude do the codebase + metrics synthesis analysis and a simple Sentry API to handle the historical metrics analysis.

Accomplishments that we're proud of

Being able to integrate all the tools in an interesting and useful way. There are so many different components that had to be setup and looked into for this to work.

What we learned

First get a baseline set of integrations working before expanding to more complex workflows. I should have first just got a simple historical analysis with Sentry API integration working first before trying to integrate Claude agentic analysis and CodeRabbit automated code reviews.

What's next for Sentralert

Built With

Share this project:

Updates