burme.sh — Burlington Mesh

The public landing page for the Burlington Mesh at burme.sh — a community off-grid MeshCore LoRa mesh for Burlington, MA. It explains what MeshCore is, lists the network's repeaters, and shows how to join the mesh.

The page itself is a single, self-contained static HTML file (inline CSS, no client-side JavaScript). It's served from a Cloudflare Worker that renders the repeater list from data/repeaters.json at request time, so adding a repeater needs no HTML edits.

Project layout

public/
├── index.html                # the page (self-contained: inline CSS, no JS)
├── burlington-mesh-logo.svg  # brand logo (favicon + hero)
├── _headers                  # security headers (CSP, HSTS, etc.)
└── robots.txt                # blocks search engines and AI/LLM crawlers
src/
└── index.js                  # Worker: serves ./public/ + renders repeaters via HTMLRewriter
data/
├── repeaters.json            # source of truth for the repeater list
├── repeaters.schema.json     # JSON Schema for the above
├── repeaters.example.json    # filled-in template
└── README.md                 # how to add a repeater
flake.nix                     # Nix dev shell: Node.js + go-task
package.json                  # Wrangler dev-dependency (installed from npm)
wrangler.jsonc                # Cloudflare Workers (Worker + static assets) config
Taskfile.yaml                 # task run / task deploy shortcuts

Adding / editing repeaters

The Repeaters section is data-driven. Edit data/repeaters.json and redeploy — the new node appears automatically. See data/README.md for the field reference and data/repeaters.example.json for a template.

Network-wide details (radio preset, RF parameters, admin contact) live in public/index.html, in the Repeaters section.

How it's served (Worker + static assets)

The site is integrated from the Worker side: wrangler.jsonc sets main to src/index.js and enables run_worker_first, so every request flows through the Worker.

• The Worker fetches the static file via the ASSETS binding (env.ASSETS.fetch(request)).
• For HTML responses it runs an HTMLRewriter pass that imports data/repeaters.json (bundled at build time) and injects the repeater cards + counts into placeholders ([data-repeaters], [data-repeater-count]). Non-HTML assets pass through untouched.

Because rendering happens in the Worker, the page ships no client-side JavaScript and keeps a strict script-src 'none' CSP. Security headers (CSP, HSTS, X-Robots-Tag, …) come from public/_headers; the page is intentionally not indexed (noindex meta + X-Robots-Tag header + robots.txt blocking crawlers).

Getting started

With Nix (recommended)

flake.nix provides Node.js 24 and go-task on Linux (x86_64, aarch64) and macOS (Intel, Apple Silicon). Wrangler is installed from npm — the dev shell runs npm install on first entry. (Wrangler is deliberately not pulled from nixpkgs, which has no prebuilt binary for Apple Silicon and would otherwise compile it from source on every nix develop.)

# Enter the dev shell (or use direnv with the included .envrc)
nix develop

Local preview

task run        # or: npx wrangler dev  — serves the Worker + ./public/ locally

Deployment

The site deploys to Cloudflare Workers (a Worker with static assets) on the custom domain burme.sh. Configuration lives in wrangler.jsonc.

task deploy     # or: npx wrangler deploy

Wrangler authenticates with your Cloudflare credentials. For non-interactive deploys (e.g. CI), set CLOUDFLARE_ACCOUNT_ID and CLOUDFLARE_API_TOKEN in the environment.