Investigation Plan: AWS Hosting + BizSite Replacement¶

Date: 2026-04-06 Author: Cathal Dempsey + Claude Status: Proposed — awaiting Phase 1 kickoff

Background¶

Current State¶

Replatform pipeline: Scrapes Wix/Mono/WP sites → Claude builds Astro pages → deploys to Cloudflare Workers/Pages
BizSite: Internal Node.js tool that pulls content from Yext/Saymore → renders JS templates to flat HTML → deploys to S3 bucket
Hosting: Split across Cloudflare (replatform) + AWS (BizSite + API server)

Problems¶

Two separate systems doing similar things (generate static sites from content)
Hosting split across two providers — two failure points
Cloudflare has a 100 project/account hard cap — need 18 accounts for 1800 sites
BizSite is a separate codebase to maintain

Proposed End State¶

One unified pipeline that can source content from either scraping (Wix migrations) or Yext API (existing BizSite clients), builds Astro sites, deploys to S3+CloudFront. BizSite retired.

Scale¶

1800 sites total
40 staff
150 sites/month migration pace (12-month Wix contract cycle)
Build cost: ~$0.30/site (Claude API)

Phase 1: Audit BizSite + Yext Integration¶

Goal: Understand what BizSite does so we know what to replicate.

Get access to BizSite repo, read the Node.js codebase
Map the Yext data model — what fields are pulled? (name, address, hours, services, photos, reviews, social links, etc.)
Document the Yext/Saymore API endpoints and auth (API key? OAuth?)
Catalogue the BizSite JS templates — how many, what page types?
Identify which of the 1800 clients currently have Yext data vs which are Wix-only
Understand the current S3 deploy — bucket structure, CloudFront setup, custom domains, SSL

Output: A mapping doc showing BizSite field → Yext field → equivalent in our content-map.json schema

Phase 2: Add Yext as a Content Source¶

Goal: The replatform pipeline currently only scrapes. Add Yext API as an alternative content source that produces the same content-map.json format.

Build lib/yext-source.js — pulls a client's Yext listing, transforms to our content-map schema
Map Yext fields to our node types:

Yext Field	Content Map Node
`name`	heading
`address`, `phone`	text / contact section
`hours`	structured data / text
`description`	text nodes
`photos`	image nodes
`logo`	header image
`socialProfiles`	socialLinks array
`reviews`	reviews array
`services`	cards / list
`categories`	nav items

Add a --source yext --yext-id <entity-id> flag to crawl.js so the pipeline can be triggered either way
Test: generate content-map.json from Yext for a client that currently has a BizSite, compare output

Output: Same content-map.json format whether content comes from scraping or Yext. Build agent doesn't need to know the difference.

Phase 3: Migrate Hosting to S3 + CloudFront¶

Goal: Replace Cloudflare Workers/Pages with AWS. Single account, no project cap.

Architecture¶

[Astro Build Output]
        |
        v
  S3 Bucket: sites.fcrweb.ie
    /waterfordcountypainters.ie/
      index.html
      assets/
      farm-painting/index.html
      ...
    /trimtech.ie/
      index.html
      ...
        |
        v
  CloudFront Distribution
    - Custom domain per client (CNAME)
    - ACM SSL cert (free)
    - Origin: S3 bucket with path prefix

Tasks¶

Audit BizSite's existing S3 bucket structure — can we reuse it or need a new setup?
Design bucket layout: single bucket with /{domain}/ prefix per client
Set up CloudFront distribution with per-client custom domains
SSL via ACM (free) — wildcard cert for *.fcrweb.ie + per-client custom domain certs

Build lib/deploy-s3.js to replace wrangler pages deploy:

aws s3 sync build/ s3://sites/{domain}/ --delete
aws cloudfront create-invalidation --distribution-id $DIST_ID --paths "/*"

DNS: clients CNAME their domain to CloudFront distribution
Test: deploy one Astro build to S3+CloudFront, verify it serves correctly

Output: node deploy.js waterfordcountypainters.ie deploys to S3+CloudFront instead of Cloudflare

Phase 4: Unified Pipeline¶

Goal: One command, two content sources, one hosting target.

CLI Interface¶

# Wix migration (scrape + build + deploy)
node orchestrate.js --url https://www.example.ie --deploy s3

# Yext client (API + build + deploy)
node orchestrate.js --source yext --yext-id 12345 --deploy s3

# BizSite replacement (bulk migrate existing BizSite clients)
node migrate-bizsite.js --all

Tasks¶

Update orchestrate.js to support --source yext flag
Build migrate-bizsite.js — iterates existing BizSite clients, pulls from Yext, builds Astro, deploys to same S3 bucket (in-place replacement)
Update dashboard to show content source (scraped vs Yext) per project
QA: side-by-side compare BizSite output vs Astro output for 5 clients

Phase 5: Retire BizSite¶

Migrate all BizSite clients to Astro builds (batched over weeks)
Verify no regressions — QA agent compares old BizSite vs new Astro for each client
Decommission BizSite codebase
Remove Cloudflare account(s) if no longer needed
Update staff training/docs for new pipeline

Cost Analysis¶

Hosting¶

	Cloudflare (current)	AWS S3+CloudFront (proposed)
1800 sites	$0 but 18 accounts	~$200-400/mo
Per-site cost	$0	~$0.15/mo
Project cap	100/account	Unlimited
Accounts needed	18	1
Static bandwidth	Unlimited	~$0.085/GB (first 10TB)
SSL	Free, auto	Free via ACM
Admin overhead	High (18 accounts)	Low (1 account)

Build Pipeline¶

Item	Cost
Claude API (builds)	~$0.30/site × 150/month = ~$45/mo
EC2 (API + scraper)	Already running
Yext API	Already paying

Total at Scale¶

Item	Monthly
AWS hosting (1800 sites)	~$300
Claude API (150 builds/mo)	~$45
EC2	Existing
Total	~$345/mo

vs maintaining two systems (BizSite + replatform) across two providers

Key Questions to Resolve¶

How many clients are on Yext today? Determines size of Phase 2 effort
Is the BizSite S3 bucket + CloudFront setup reusable? Could skip most of Phase 3
Does Yext have rate limits? Matters for bulk migration
Do BizSite clients have custom domains on CloudFront already? If yes, migration is just swapping HTML files — zero DNS changes
Can we get read access to the BizSite repo? Needed for Phase 1
Are there BizSite features beyond content rendering? (analytics, tracking, integrations we'd need to replicate)

Timeline Estimate¶

Phase	Duration	Dependencies
Phase 1: Audit	1 week	BizSite repo access
Phase 2: Yext source	1-2 weeks	Phase 1 complete
Phase 3: S3+CF hosting	1 week	Can run parallel to Phase 2
Phase 4: Unified pipeline	1 week	Phases 2+3 complete
Phase 5: BizSite retirement	4-8 weeks	Phased migration