Back in February, I spec’d a new PC to dive into AI/LLMs and sharpen my existing networking skills. By May, I’d spun up a full observability stack, built version control, hosted internal DNS, deployed Docker and LXC services, and explored a suite of self-hosted tools — with ChatGPT as my hands-on assistant and research partner throughout.
What began as a modest AI lab has evolved into a production-grade home infrastructure — reconnecting me with my sysadmin roots while pushing into automation, AI integration, and modern networking.
Next up: populating NetBox as the single source of truth for my network, laying the groundwork for LLM-powered querying and infrastructure automation.
Here’s a snapshot of what I’ve built so far, and a preview of the topics I’ll dive into in upcoming posts.
Work Completed
Lab Rig
- Chosen Spec specifically to support a medium LLMs
- Dual-boot Linux/Windows
- Geforce 4070 Ti Super
- 128GB RAM
- Containerlab installed
- First tests with LLM images completed
OpnSense Firewall
- Rebuilt a 2016 gaming PC into a dedicated OpnSense box
- Configured WireGuard VPN
- Added 10G NIC to Firewall to Juniper EX2300
- Fully migrated from ISP CPE to OpnSense
Mon1 – Docker Monitoring Host
- Dockerized:
- Grafana
- Prometheus
- cAdvisor
- Portainer
- Node_Exporter
- Junos_Exporter
- InfluxDB
- Technitium DNS (DNS1)
- GitLab
- ELK Stack: Elasticsearch, Logstash, Kibana, Filebeat, Elastalert
- Maltrail, Zenarmor Cloud, CrowdSec
- Home Assistant metrics exported to InfluxDB, ready for dashboards
- IDS enabled on OpnSense
- Built Mon1 dashboards in Grafana
- Temporarily crashed due to RAM pressure from GitLab + Security stack
- Upgraded to 32GB RAM, resumed services
- Slack & Telegram alert testing with Grafana
- Added Homarr dashboard
- Backup scripts using Python + crontab
