Automating Coding Agents To Accelerate Issue Triage

CI pipeline unexpectedly broken?
Unsure what is causing the sudden influx of runtime errors after a release?
Sudden increase in user response times with no obvious cause?

Agentic coding assistants can help with all of these issues. In your IDE/editor as an individual contributor ***OR ***as a standalone component of your DevSecOps pipeline.

By integrating an agent into your pipeline that is tailored to how you or your team works, you can save precious time on initial issue triage, allowing your team to begin working on corrective measures or resolving simple issues as soon as they appear.

Introducing agents into your workflow isn't intended to replace any critical work; it's meant to accelerate remediation of quick fixes and help team members avoid starting from square one when tackling a critical issue.

In this talk, Travis will show you how to build a simple GitHub action triage agent —one that investigates and attempts to fix issues when a GitHub action fails. As we build our triage agent, we'll discuss which parts of the workflow are (and aren't) good candidates for handoff to an LLM, depending on whether the necessary information is easily accessible in the CI environment, via an API, or through another pre-existing solution.

By relying on what we know and letting the agent focus on what we don't, we can ensure that our LLM-powered solutions are set up for success and replicate the steps a member of our team would take during initial triage of an issue.

About Our Speaker:

Travis is a Field Engineer at Sourcegraph, focused on making sure that current and future customers are able to maximize what’s possible with today’s frontier models using Amp, Sourcegraph's agentic coding assistant.