SUMMARIZATION

Rollback recommendation when a deploy spikes errors

When a Sentry alert fires for an error spike, attributes it to the most recent deploy, summarizes whether the spike is dominated by new error classes introduced by that release.

CategorySummarization
Enginesim
Difficultyadvanced
Triggerwebhook
Steps6
Setup~25 min

How it runs

The automated pipeline, trigger to output.

  • TriggerSentry error-spike alert webhookSentrySentry
  • ActionIdentify the most recent deployVercelVercel
  • ActionFetch issues driving the spike and tag originSentrySentry
  • LogicCompute share of spike from the new release
  • ActionWrite a go/no-go rollback recommendationOpenAI
  • OutputPost recommendation to the incident channelSlack

What it does

This workflow turns a raw error-spike alert into a rollback decision. When Sentry reports a spike, it finds the deploy that shipped just before it, pulls the issues driving the spike, and determines how much of the volume comes from error classes that release introduced versus pre-existing noise. An LLM weighs that and writes a clear recommendation: roll back, or hold and investigate, with reasoning.

When to use it

Use it during high-traffic launch windows or for teams without a dedicated SRE on every shift. It compresses the panicked "is it the deploy?" investigation into one Slack message with a defensible recommendation and the exact deploy to revert.

How it works

A Sentry metric-alert webhook fires on the spike. The flow identifies the latest deploy and fetches the issues contributing to the spike, separating new-from-this-release from carried-over. A logic step computes the share of spike volume attributable to the new release. An OpenAI step summarizes the situation and emits a go/no-go rollback call. The recommendation, with the deploy link and contributing issues, is posted to the incident Slack channel.

Set it up

What you configure once, before turning it on.

  1. 1
    Connect SentryErrors, performance, releases.
  2. 2
    Connect VercelDeploys, runtime logs, analytics.
  3. 3
    Connect OpenAIModels, embeddings, files.
  4. 4
    Connect SlackChannels, DMs, threads, mentions.
  5. 5
    Set each agent's modelWe leave models unset so you pick the tier — fast + cheap, or top-quality.
  6. 6
    Tune it to your dataEdit the prompts, filters, and field mappings so it matches how your team works.
  7. 7
    Test, then turn it onRun once against a sample, confirm the output, then enable the trigger.

Run this workflow in your colony.

14-day trial. No DevOps. No Sales call. Provisioned in under a minute.