Can AI replace HitKeep Cloud?

Replace HitKeep Cloud by deploying the real open-source HitKeep, not by rewriting analytics.
Use Docker Compose and the official pascalebeier/hitkeep image; do not build a clone.
Create compose.yml for one HitKeep service with restart: unless-stopped.
Persist the data, retention archive, and backup snapshots in separate named volumes.
Put HITKEEP_PUBLIC_URL and HITKEEP_JWT_SECRET in .env and ship a safe .env.example.
Generate the signing secret locally; never commit it or print it into documentation.
Bind HitKeep to localhost on port 8080 so my existing reverse proxy can terminate TLS.
Document the exact reverse-proxy requirements for the public analytics domain.
Enable automatic local backup snapshots and keep a bounded retention window.
Add backup.sh to copy a consistent snapshot to a second destination and document restore.
Add update.sh to pull the latest image, recreate the service, and show recent logs.
Include a health check that uses the official image's supported health-check command.
Keep optional SMTP, social login, Search Console, MCP, and AI provider settings disabled.
Do not add PostgreSQL, Redis, ClickHouse, Kafka, or a separate queue service.
Do not add accounts, billing, telemetry, or a hosted control plane around HitKeep.
Out of scope: modifying HitKeep, rebuilding its dashboard, or promising managed uptime.
Write a README covering DNS, TLS, first-owner setup, adding a site, and the tracker tag.
Document where every persistent file lives and how to test a restore before trusting backups.
Start the stack, verify the container is healthy, and confirm the dashboard and hk.js load.
Finish by listing the exact verification commands and the remaining operations I now own.