Postmortem: incident log + postmortem writer in Bolt

Small engineering teams usually keep incidents in a Slack thread and write the postmortem a week later from memory. Postmortem keeps both in one place: log the incident while it is happening, and the write-up is already done when it is over.

Dashboard with open incidents, median time to detect, mitigate and resolve over the last 30 days, and incidents per week for the last 8 weeks.

Each incident has a timeline, five-whys root cause, action items with owners and due dates, and timestamp validation so detected can never come before started.

The postmortem renders live from the data as Markdown: summary, impact, timeline, detection, root cause chain, action items. Copy it or download it as a dated .md file.

Local-first: everything stays in the browser, no account, no backend.

One prompt in Bolt, written as a spec: data model, derived fields, four routes, design tokens and validation rules. Bolt built it in one pass with no TypeScript errors.

Vite + React + TypeScript + Tailwind, react-router, a plain SVG chart, and pure helpers for time and markdown so they can be unit-tested.

Design: dark operational theme, tabular numerals, monospace timestamps, 1px borders, no gradients.

The sample incidents in the images are tagged SAMPLE inside the app. They show the layout; they are not real outages.

Live: https://postmortem-incident-w8tg.bolt.host