Blackrack

Blackrack: Your server, your data. Managed self-hosting on k3s replacing SaaS with private personal clouds. AI-orchestrated infrastructure for total data sovereignty. Stop renting, start owning

Comment

Inspiration In an era where "the cloud" is just someone else's computer, we’ve become digital tenants. We rent our memories to Google Photos, our documents to iCloud, and our privacy to whoever offers the most convenient "free" tier. Blackrack was born from a simple, defiant realization: You shouldn't have to be a full-time sysadmin to own your data.

The inspiration was digital sovereignty. We wanted to build a bridge between the complexity of self-hosting and the convenience of SaaS. Our goal was to create a "black rack" of infrastructure—invisible, secure, and entirely yours—that replaces expensive subscriptions with local ownership and privacy.

🛠 What it does Blackrack is a managed self-hosting infrastructure service. We provide technical or curious clients with their own private server environment (a "gateway") that runs a curated stack of open-source alternatives to big-tech services.

  • Data Sovereignty: Migrate from Google Drive, Dropbox, and iCloud to tools you control.
  • Personal AI Assistants: Every gateway includes a private assistant powered by our custom OpenClaw and Hermes runtimes.
  • Hybrid Infrastructure: Deploy on high-availability cloud (AWS) or your own local hardware (Proxmox/Raspberry Pi) while maintaining a unified management layer.
  • Automated Ops: We handle the "boring" parts—backups, security patches, SSL certificates, and monitoring—so you can just use your tools.

🏗 How we built it We architected Blackrack as a distributed, multi-tenant Kubernetes (k3s) cluster.

  • Orchestration: We use k3s across a hybrid topology of AWS EC2 nodes (for reliability) and Bogotá-based Proxmox workers (for local cost-efficiency).
  • Storage: Longhorn provides block storage with 3-way replication, ensuring that even if a node disappears, your data doesn't.
  • Connectivity: Tailscale creates a secure VPN mesh, allowing our Bogotá and NYC nodes to communicate as if they were in the same room.
  • AI Intelligence: We built a centralized LiteLLM proxy to manage high-quality reasoning models like $MiniMax-M3$, coupled with a custom provider-broker that isolates sensitive API credentials from client runtimes.
  • Deployment: A fully automated CI/CD pipeline using GitHub Actions allows us to roll out fleet-wide updates with a single commit.

🌋 Challenges we ran into Building a "cloud at home" isn't easy. Our biggest challenge was The Great Migration. We started on a single Docker host but quickly outran its limits. On May 26, 2026, we "destroyed" our legacy host and migrated 20+ runtimes to Kubernetes in a single weekend.

We also battled The WAN Latency Demon. Attempting to replicate storage (Longhorn) between Colombia and the US over a 150ms RTT link proved impossible for synchronous quorums. We had to pivot to a "two-tier" storage strategy: high-durability cloud for paying clients and a high-performance local-path tier for our family/dev cohort.

Finally, we faced the Spot Node Paradox. To keep costs low, we used AWS Spot Instances. However, an "ungraceful reclaim" once took the entire fleet offline. We had to build an automated "Never-Again" guard—a Spot Node Verifier—that health-checks replacement nodes before allowing them to join the "living" fleet.

🏆 Accomplishments that we're proud of

  • Fleet Scalability: We can now provision a complete, secure client environment in under 15 minutes using our provision_client_pipeline.py.
  • Credential Isolation: We successfully implemented a security model where client gateways never hold raw provider keys, preventing a single compromised pod from leaking our global infrastructure secrets.
  • Agentic Hub: We developed a "Hub-and-Spoke" agent model where a central COO agent coordinates specialized roles (Ops, Sales, Legal) to manage the company autonomously.
  • The 10k Starter: We launched a $10,000 COP/mo (~$2.50 USD) entry tier to prove that self-hosting can be cheaper than a coffee.

📚 What we learned We learned that Infrastructure is Code, but Operations is Documentation. In a distributed system, a change in one place can have ripples 2,000 miles away. We adopted a "Global-First Fix" discipline: if one client's assistant misbehaves, we don't patch the client; we patch the Baseline ConfigMap so the whole fleet gets smarter.

We also learned the LaTeX of ROI. We realized that for a client, the ROI of Blackrack can be expressed as: ROI = (∑ (SaaS_Costs) - (Blackrack_Subscription))/(Total_Data_Ownership + Privacy_Peace_Of_Mind) If Privacy → ∞, then ROI → ∞, regardless of the monthly cost.

🚀 What's next for Blackrack Our immediate goal is to reach our "Break-Even" target of 8 paying customers.

Technically, we are working on Blackrack ID: a unified SSO experience (via Authentik) where one set of credentials unlocks your entire self-hosted world. We are also refining our Hermes agentic framework to make our assistants more proactive, helping users manage not just their data, but their digital lives.

Your server, your data. Welcome to Blackrack.

Built With

Share this project:

Updates

Submission history