seo-auditor

A ProAgentStore agent that audits website SEO daily and tracks score trends over time.

What it does

• Fetches each configured URL at 05:00 UTC daily
• Extracts: title, meta description, H1-H6 structure, image alt tags, internal/external links, word count, JSON-LD schema markup
• Calls Workers AI (llama-3.3-70b) to score each page 0-100 and generate actionable recommendations
• Falls back to deterministic rule-based scoring if AI response is malformed
• Persists each audit to D1 and a rolling 90-entry score history to a Durable Object per site
• Detects regressions (score dropped vs previous audit) and logs a warning

API

Method	Path	Description
POST	/sites	Register a URL for daily auditing
GET	/sites	List all registered sites
GET	/audits	Paginated audit list (latest first)
GET	/audits/latest	Latest audit per site
GET	/audits/:id	Single audit by id
POST	/cron/trigger	Manually run a full audit cycle
GET	/health	Liveness probe

POST /sites

{
"url": "https://example.com",
"label": "Example homepage"
}

GET /audits

Query params: limit (max 100, default 20), offset (default 0), site_id (filter to one site).

Audit object

{
"id": "abc123def456",
"site_id": "xyz789",
"audited_at": "2026-06-06T05:00:00.000Z",
"score": 74,
"title": "Example Domain",
"meta_desc": "This domain is for use in illustrative examples.",
"word_count": 312,
"h1_count": 1,
"images_total": 4,
"images_no_alt": 1,
"links_internal": 8,
"links_external": 2,
"has_schema": true,
"recommendations": [
"Expand meta description to 120-160 characters (currently 48).",
"Add alt text to 1 image that is currently missing it.",
"Expand content beyond 600 words for stronger SEO signals."
],
"regression": false
}

Scoring

The AI scores pages on these criteria (rule-based fallback uses the same weights):

Signal	Points
Title present	10
Title 50-60 chars	10
Meta description present	10
Meta description 120-160 chars	10
Exactly one H1	15
300+ words	10
600+ words	5 bonus
All images have alt text	10 (−2 per missing, max −10)
JSON-LD schema present	10
Internal + external links	10

Development

pnpm install
pnpm dev

Create the D1 database locally:

wrangler d1 create seo-auditor-db
# copy database_id into wrangler.toml
wrangler d1 migrations apply seo-auditor-db --local

Deploy

Push to main — GitHub Actions runs migrations then deploys the worker.

Or manually:

pnpm run db:migrate
pnpm deploy

Secrets

Set via wrangler secret put and mirror in Doppler project pags:

Secret	Purpose
ADMIN_TOKEN	Bearer token for write API endpoints (optional; if unset all routes are open)