AI Incident Commander

AI Incident Commander is a simple on‑call assistant that turns raw system alerts into a clear, human‑readable incident story. It opens incidents, summarizes what’s happening, suggests what to do next, and captures a post‑incident summary.

What it does (plain English)

• Watches for problems (errors, latency spikes, crashes).
• Creates a new incident automatically when something breaks.
• Summarizes the symptoms and highlights likely causes.
• Suggests next steps using runbooks.
• Drafts an incident summary for postmortems.

Why it’s valuable

• Saves time during incidents by reducing noise.
• Gives non-experts a clear, readable incident overview.
• Provides a repeatable demo of “AI + ops” that’s easy to show in a portfolio.

How it works (simple flow)

1. A demo service fails (simulated fault).
2. The backend records the incident and timeline events.
3. An analysis pipeline creates a summary and stores it in S3.
4. The React dashboard shows the incident list and timeline.

What you can see in the demo

• A live incidents list with severity, service, and status.
• A timeline of events and analysis updates.
• A “Request Analysis” button that triggers the workflow.
• Runbook recommendations filtered by service/tag.

Architecture (high level)

• Frontend: React dashboard hosted on S3 + CloudFront.
• Mobile: Flutter app for alerts and incident summaries.
• Backend: Node.js Lambdas behind API Gateway for incidents/runbooks/comments.
• AI/Analysis: Python Lambda (stub) invoked via Step Functions, stores summaries in S3.
• Workflow: EventBridge for incident events, Step Functions for analysis pipeline.
• Data: DynamoDB for incident state, S3 for logs/summaries, optional OpenSearch for search.
• Infra: CDK stacks, EKS demo services with OpenTelemetry and chaos testing.

Repo layout

• apps/web: React dashboard.
• apps/mobile: Flutter alerts app.
• services/api: Node.js API Lambdas.
• services/analysis: Python analysis Lambdas.
• infra: AWS CDK infrastructure.
• docs: Architecture, data model, API, and demo flow.

Documentation

• docs/ARCHITECTURE.md
• docs/DATA_MODEL.md
• docs/API.md
• docs/DEMO.md

Deploy script

From the repo root:

./scripts/deploy.ps1

Optional:

./scripts/deploy.ps1 -ApiBaseUrl https://your-api-id.execute-api.region.amazonaws.com
./scripts/deploy.ps1 -SkipWebBuild
./scripts/deploy.ps1 -SkipWebDeploy

The script deploys CDK infra, builds the web app, and syncs apps/web/dist to the frontend S3 bucket (with a CloudFront invalidation).

Status

Core API + dashboard + analysis stub are working. Next steps: replace analysis stub with Bedrock, expand runbook management, and add real observability signal ingestion.