Infrastructure

The Canadian Space runs on a self-hosted setup. We own the hardware contract, manage the deployments, and see exactly what's happening at every layer. Here's why that matters, and how it's built.

The layout

At the core is a single OVH Cloud VPS (VPS2 tier — 6 vCores, 12GB RAM, 100GB NVMe, Frankfurt) running Docker Compose. Inside: n8n (workflow orchestration), Redis (job queue), Caddy (reverse proxy + TLS), and a handful of Python scripts we invoke over SSH. Around it: WordPress on Bluehost, GitHub for code and image hosting, OpenRouter for LLM routing, and a handful of aerospace data sources feeding stories in.

graph TB
    Data["📡 <b>Data sources</b><br/>Spaceflight News API · Launch Library 2 · RSS<br/>X (via Rettiwt-API) · News sites (via Crawl4AI + CRW self-hosted)"]

    subgraph OVH["🖥️ OVH Cloud VPS — self-hosted core"]
        Core["⚙️ <b>n8n</b> workflow engine<br/><small>every pipeline runs here</small><br/><br/>🔒 Caddy &nbsp;·&nbsp; 📦 Redis &nbsp;·&nbsp; 🐍 Python scripts"]
    end

    LLM["🤖 <b>LLMs via OpenRouter</b><br/>Qwen 3.7 Plus (author) · Claude Haiku 4.5 (fallback)<br/>GPT-5-mini (editor) · Grok (social copy)"]

    WP["📡 <b>WordPress on Bluehost</b><br/>thecanadian.space"]
    Social["📱 <b>Facebook + Instagram</b>"]
    GH["💾 <b>GitHub</b><br/>tcs-images · workflow backups · repos"]

    Data -->|feeds| Core
    Core <-->|LLM calls| LLM
    Core -->|publishes| WP
    Core -->|distributes| Social
    Core -->|persists| GH

    classDef input fill:#0A1428,color:#fff,stroke:#0A1428
    classDef ai fill:#FF9D3D,color:#000,stroke:#FF9D3D
    classDef out fill:#22C55E,color:#fff,stroke:#22C55E
    classDef store fill:#334155,color:#fff,stroke:#334155
    class Data input
    class LLM ai
    class WP,Social out
    class GH store

Legend: dark blue = data in · orange = AI · green = published out · slate = storage · orange-bordered box = self-hosted OVH core

Why self-hosted?

We made a deliberate choice to self-host instead of using a fully managed platform (Zapier, Make, etc.). Here's why:

Control. We see exactly what we're running and control our own destiny. No surprise pricing tiers, no "your feature request is on the roadmap," no waiting for a third party to approve a new data source.

Learning. Running our own infrastructure keeps us sharp. We understand caching, queueing, error handling, and production operations — not just the happy path.

Openness. Open-source at the core (n8n, Docker, Caddy) means anyone can audit what we're doing, contribute improvements, or fork and build their own version.

Disaster recovery & redundancy

Our n8n workflows are version-controlled on GitHub (auto-backup hourly). WordPress backups are automated by the host. If the VPS goes down, we can spin up a new OVH instance and restore from the backups in under an hour.

For critical workflows (Daily Broadcast, editorial routing), we've built fallback routes: if Qwen errors mid-response, Claude Haiku 4.5 steps in. If one data source is down, the others keep pulling. The pipeline degrades gracefully rather than failing outright.

---

Want more?

See the full component-by-component breakdown on Tech Stack.