Overview
Deterministic Kafka replay for debugging event windows safely.
Lighthouse is being rebuilt from a legacy monitoring dashboard into a Kafka replay workbench. The current implementation already supports a local sample cluster, external Kafka and Confluent Cloud connection settings, a Prometheus-backed dashboard, and a replay CLI with dry-run preview, timestamp-window replay, throttling, and replay traceability headers.
Run locally
Start the full demo stack with Kafka, Prometheus, and the app.
npm run docker:sample
Preview a replay
Inspect a bounded slice without producing anything.
npm run replay:cli -- --source orders --destination orders-replay --partition 0 --start 10 --end 25 --dry-run
Run a replay
Copy one closed offset range or resolved time window into a sandbox topic.
npm run replay:cli -- --source orders --destination orders-replay --partition 0 --start 10 --end 25
What Exists Today
- Prometheus-backed Kafka dashboard in Next.js
- Local three-broker Kafka sample stack in Docker Compose
- Kafka metrics exporter for local Kafka and Confluent Cloud
- Replay CLI with offset-range and timestamp-window replay, throttling, dry-run preview, and metadata headers
- Replay job API for persisted replay creation, preview, and start
- Replay workspace in the application shell for draft creation and monitoring
- CI for audit, lint, unit tests, build, Docker, and Playwright
Read This First
Getting Started
Choose between the sample cluster, an existing Kafka endpoint, or Confluent Cloud.
Replay CLI
See the command contract, dry-run mode, and replay header model.
Replay Jobs
Use the persisted workflow for draft, running, completed, failed, and cancelled jobs.
Replay API
Create, preview, start, cancel, and inspect replay jobs through HTTP endpoints.
Replay UI
Use the browser workflow for draft creation, preview, and job monitoring.
Architecture
Understand the current dashboard flow and the replay execution path.
Operations
Use the runbook for health checks, troubleshooting, and local verification.
MVP Acceptance
Review the end-to-end checks that prove the replay MVP before alpha hardening.
site/ and the workflow publishes it
on pushes to main or by manual dispatch.