🧪 AI Lab VPS

Learning · Building · Publishing (GitHub + HF)

A personal AI engineering laboratory on an OVHcloud VPS, the sandbox where I learn Linux administration, Docker, backend development and agentic AI by building real projects. Each one gets published publicly as a portfolio.

What's here

• roadmap.md — master plan: infrastructure Phases 0–4, a 5-month learning roadmap, backup strategy and a services/access matrix
• journal.md — personal journal, one entry per day
• project-conception-log.md — timeline of every decision, chosen or rejected, and why
• achievements.md — running record of GitHub achievements earned by this account
• rescue-drill.md — rescue-mode runbook (tested 2026-08-06)
• assets/ — project beaker icon (SVG + PNG exports, favicon)
• hello/ — FastAPI "hello" app (containerized in Phase 3)
• scripts/ — backup.sh (age-encrypted backups to Backblaze B2, nightly via systemd timer)
• .opencode/ — opencode config, including the /end-session command
• print.css + ai-lab-summary.pdf — printable PDF of the plan

Infrastructure status

Phase	Status
0 — Prep	✅ complete
1 — Server Foundation	✅ complete (SSH hardening, UFW, edge firewall, Fail2Ban)
1.5 — Systemd-first	✅ mostly complete
2 — Docker	✅ complete
3 — Slim Stack (Postgres 17 · Caddy · FastAPI)	✅ complete
4 — Backup & Resume (Backblaze B2)	✅ complete

Philosophy

• Systemd-first — learn Linux before Docker
• Slim stack — only what's being studied now, nothing exposed but 80/443
• Secrets in Infisical — no .env on the server, no secrets in git
• Publish everything — code → GitHub, demos → Hugging Face, config → Infisical, data → encrypted backups

License

MIT © 2026 Bruno Lima