Sentry Incident Investigation Workflow with Gemini, Slack & Postgres

Turn Sentry alerts into structured incident investigations—automatically

This Sentry Incident Investigation Workflow with Gemini, Slack & Postgres listens for Sentry issue alert webhooks, enriches each incident with Sentry + GitHub context, uses Google Gemini to generate a JSON investigation report, saves everything to Postgres, and posts a clear summary to Slack.

What this workflow does

• Receives Sentry issue alerts: Accepts a POST webhook from Sentry and normalizes the payload into a consistent incident context (including issue ID, title, environment, and relevant links).
• Prevents duplicate investigations: Checks Postgres for an existing record using the same Sentry issue ID, proceeding only if it’s new.
• Enriches incident evidence: Fetches the latest issue details from Sentry and pulls recent commits from the configured GitHub repository.
• Finds related history: Searches Postgres for up to 10 similar historical incidents based on the issue title and prior incident reports.
• Generates a structured investigation: Bundles Sentry data, GitHub commit context, and historical incidents, then sends the evidence to Google Gemini to produce a JSON report with severity, hypotheses, and recommended actions.
• Validates, scores, and stores results: Validates and scores the AI output, then upserts the investigation (evidence + report) into Postgres.
• Communicates to Slack: Posts an incident summary to a Slack channel including severity, confidence, and a link back to the Sentry issue.

Use cases

• Reduce time-to-triage by auto-drafting investigation reports for every new Sentry issue.
• Enable faster incident response with consistent summaries posted directly to Slack.
• Leverage historical incident patterns stored in Postgres to inform Gemini’s hypotheses.

Technical details

• Webhook: Sentry issue alert webhook ingestion
• if / Code: Normalize payload and control “already processed” logic
• Sentry integration: Fetch latest issue details
• GitHub integration: Retrieve recent commits
• Postgres: Check duplicates, search similar incidents, upsert evidence + investigation report
• Google Gemini: Generate a structured JSON investigation report
• Slack: Post severity/confidence summary with a Sentry link