Skip to content
SWEny

SWEny Triage

SWEny Triage is the first recipe built on the engine. It automates the on-call triage loop: monitor your observability logs, identify the highest-impact novel issue, create a ticket with root cause analysis, implement a fix, and open a PR — all without human intervention.

How it works

Section titled “How it works”

Learn

Section titled “Learn”
  • Queries your observability provider (Datadog, Sentry, CloudWatch, Splunk, Elasticsearch, New Relic, or Grafana Loki) for errors in the configured time window
  • Aggregates errors by service and pattern
  • Searches your issue tracker (Linear, GitHub Issues, or Jira) for existing tickets
  • Checks open PRs to avoid duplicate work
  • Ranks issues by impact and novelty

Act

Section titled “Act”
  • Investigates the top novel issue using Claude AI — reads error logs, traces root causes through the codebase
  • Creates a detailed ticket with root cause analysis
  • Creates a branch, implements a fix, and opens a PR for human review
  • Links the PR back to the originating issue

Report

Section titled “Report”
  • Posts a summary to GitHub Actions, Slack, Teams, Discord, email, or webhooks
  • Includes links to the ticket and PR, investigation details, and skipped items

Key features

Section titled “Key features”

Novelty gate — Triage won’t create duplicate tickets. It checks your issue tracker and open PRs for existing issues before acting. Known issues get a “+1 occurrence” comment instead.

Cross-repo dispatch — If the bug belongs to a different repository (determined via your service map), Triage automatically dispatches the fix workflow to the correct repo.

Dry-run mode — Run the full investigation without creating tickets or PRs. Useful for testing your configuration before going live.

Configuration

Section titled “Configuration”

Triage is configurable through GitHub Action inputs or programmatically via the engine:

  • Inputs — time range, severity focus, service filter, investigation depth, provider credentials
  • Outputs — issues found, recommendation, issue URL, PR URL
  • Examples — dry runs, specific issue targeting, service filtering

Get started

Section titled “Get started”

Follow the Quick Start to set up Triage in your repo in 5 minutes, or see the End-to-End Walkthrough for a real scenario from error spike to merged PR.