What I have built and when, across the Beelab platform, the AI layer, and the site. Updated as releases go out.
Local LLM inference across an Apple Silicon mesh on a Tailscale network. A vendor-neutral router fronts the cluster with model-name routing and automatic fallback, RAG runs on Qdrant and pgvector, and you can bring your own keys for optional cloud burst.
Hardware sourcing now covers concierge install, OEM drop-ship, and the inventoried Beelab Box. Spec floor: Intel vPro with Intel I219, I225, or I226 NICs across HP EliteDesk, Lenovo ThinkCentre, Dell OptiPlex, Minisforum, and Apple Silicon Mac mini Pro or Mac Studio.
Self-hosted app catalog wired across all VMs, 80+ apps you pick, anything containerizable. Defaults ship pre-configured and any layer stays swappable for your own choice.
Multi-path backups built into the platform: hypervisor image snapshots, encrypted file-level backups, an on-prem mirror, and scheduled offsite to a provider you pick, with health-check monitoring on every job.
The platform ships a commit-to-deploy pipeline: lint, test, vulnerability scan, build, sign, push, policy check, deploy, and continuous drift reconciliation. Artifact signing with a software bill of materials (planned v3.0).
Networking wired for zero trust: a Tailscale mesh for private access, VLAN segmentation, and a hardware firewall at the edge. The default stack swaps for your own firewall or tunnel.
Production runs on Talos immutable Kubernetes, with k3s for lab and edge, Proxmox for virtualization, and Docker for container workloads, coexisting side by side. Scales from one node to N.
Reorganized the marketing site from one long scroll into dedicated pages for Architecture, AI, Apps, Hardware, Services, Pricing, and this Changelog. Supporting pages (Specs, Roadmap, Docs, Community) surface from the footer.
Updated as changes ship. Older entries are trimmed periodically.