GitHub

Automated Postmortems

Operyn automatically generates comprehensive postmortem reports after an incident is resolved. This feature leverages LLMs to synthesize incident data, logs, and metrics into a structured Markdown report.

How it Works

  1. Trigger: When an incident's status is updated to RESOLVED, the Incident Engine triggers the postmortem generation workflow.
  2. Context Gathering: The system collects:
    • Incident details (title, description, severity, affected services).
    • AI Diagnosis results (root cause, suggested fixes).
    • Timeline of events and remediation actions taken.
  3. AI Generation: The data is sent to the AI Engine's PostmortemAgent.
  4. Agent Workflow:
    • The agent uses a structured graph (via LangGraph) to process the context.
    • It adheres to a standardized postmortem template.
    • It performs automated validation of the generated Markdown.
  5. Persistence: The generated report is saved to the database and associated with the incident.

Postmortem Template

Every generated postmortem includes:

  • Executive Summary: A high-level overview of the incident.
  • Impact Analysis: Detailed assessment of which services and users were affected.
  • Root Cause Analysis (RCA): Deep dive into the technical cause of the failure.
  • Timeline: Chronological list of events from detection to resolution.
  • Remediation Actions: What was done to fix the issue.
  • Action Items: Recommended follow-up tasks to prevent recurrence.
  • Operyn Intelligence Recommendations: Structured follow-up guidance grouped into:
    • Policy Improvements
    • Runbook Updates
    • Prevention Actions

In the incident postmortem panel, these recommendations are paired with direct action links to:

  • Runbook workspace (/console/runbooks)
  • Remediation policy workspace (/console/remediation)

The intelligence strip also supports:

  • One-click draft runbook creation from recommendation content.
  • Copying a prefilled policy patch proposal before jumping to remediation controls.
  • Draft creation redirects to runbooks with status=draft preselected for immediate review.
  • Policy patch handoff now passes policyPatch via remediation URL, auto-opens policy configuration, and prefills editor context.
  • Policy editor now includes Apply suggested approvals to parse recommendation text into manual-approval actions and approval environments.

Configuration

Postmortem generation can be customized in the AI_ENGINE_CONFIG:

  • LLM Provider: Choose between OpenAI, Anthropic, or Gemini.
  • Template Context: Configure how much log/metric data is included in the prompt.