Our Technology

Built on the
modern edge stack

Spif.io combines Vite + Svelte compilation with Cloudflare's global edge network: pages answer from the edge in milliseconds and score 85-86 on Lighthouse, measured on our production template sites, capture-dated and published. That's a lab score. Lab is gameable; real-user field data isn't - we publish both, wins and losses.

The architecture · drawn from production

How it all connects

Feed drop to edge delivery: every terminal, trace, and cadence below is the system as it runs today - the pooled SFTP drop, the three-layer truth core, and the Cloudflare edge your pages answer from.

HOMENETDEALERCENTERDEALERON*.csvSFTP DROPone ssh user / providerrouted by dealer idDROP-WATCHERFEEDS PROCESSORR2 ARCHIVEevery drop archivedcontent-addressed · 90 dabsence in a drop= same-cycle removalTRUTH COREFACTSwhat the feed said▸ INTENTSwhat humans asked▸ REFEREE arbitrates writes▸ PROJECTIONthe vehicles tableHUMANEDITSenter asintentsprojection · vehiclesCLOUDFLARE EDGEPAGES · PRERENDERWORKER APIHYPERDRIVE ▸ POSTGRESrebuilds fan out on writeDEALER SITESMARKETPLACEMERCHANT FEEDSgoogle merchant apispif.io · inventory pipeline · drawn from production
The edge platform behind those terminals
Why Cloudflare?

The unfair advantage for dealership sites

We build on Cloudflare's global edge network and leverage every feature of it to make your dealership website fast - the platform's own published specs below.

300+

edge data centers

your site answers from the closest one, whether the shopper is in Miami or Montana

<50ms

KV reads, worldwide

vehicle listings, pricing, and dealer config cached at the edge, no origin round-trip

Share of the web served by Cloudflare 20%
Workers cold starts 0ms
Cache hit rate at the edge up to 95%
Network capacity absorbing attacks 100 Tbps+
Global cache purge on inventory change <30s
SSL/TLS certificates, every page included
Traffic + performance analytics real-time

platform figures - Cloudflare's published specifications, not our measurements. Ours carry capture dates: the evidence board above and /compare.

The stack, component by component

Component by component

The complete technology stack

Every component chosen for speed, reliability, and developer experience House rule: 0 = null. Most feeds write "0" for a missing price - the pipeline treats it as missing, never as a real number.

Frontend Compilation

Vite

Build tool with instant HMR and optimized production builds

ES modules

Tree shaking

Code splitting

Asset optimization

Svelte

Compiles to vanilla JS with zero runtime overhead - no virtual DOM

Reactive declarations

Compile-time optimization

Minimal bundle size

Built-in animations

Edge Infrastructure

Cloudflare Pages

JAMstack platform for deploying static sites with unlimited bandwidth

Git integration

Preview deployments

Automatic builds

Global CDN

Cloudflare Workers

Serverless functions running at 300+ edge locations with 0ms cold starts

V8 isolates

Sub-millisecond startup

Unlimited scale

Global deployment

Data & Storage

Cloudflare KV

Distributed key-value store with eventual consistency

Global replication

<50ms reads

Billions of keys

Automatic expiration

Cloudflare D1

SQLite at the edge for structured data that needs relational queries

SQL queries

Edge-local reads

Strong consistency

Point-in-time recovery

Pipeline & Analytics

Feed Processing Pipeline

Multi-provider inventory ingestion with every drop archived and monitored end to end

HomeNet & DealerCenter adapters

Content-addressed feed archival

Out-of-band anomaly monitor

Automatic site rebuilds

Analytics & Attribution

Automotive-standard GA4 events with per-vehicle lead attribution

ASC GA4 event spec

VIN-level lead attribution

Core Web Vitals tracking

Search Console integration

Ready to see it in action?

Request a demo and we'll stand up a staging site on your real inventory, or measure the site you have today, script off vs on.