# Pure Light Ltd: Edge Execution Chassis, engineered properly
> UK software company operating one unified Edge Execution Chassis: autonomous technical SEO (Cleo SEO), continuous compliance (AuditShield), and AI agent governance (PARLEY), fail-open at the CDN boundary.

Pure Light Ltd (Company No. 17433578) designs, builds, and operates a composable edge execution chassis on the Cloudflare network. Modules execute at the CDN boundary via Rust HTMLRewriter stream processing, Durable Objects, and D1, every layer fail-open by design: if the edge ever hiccups, origin traffic passes unimpeded.

- Legal Operator: Pure Light Ltd (Registered in England & Wales, No. 17433578)
- Registered Office: 66 Paul St, London EC2A 4NA, United Kingdom
- Leadership: G. Butterworth, Founder & Director (UK Government Enhanced DBS Certified)
- Institutional Indemnity: £5,000,000 ($5M) Professional Indemnity + £5,000,000 ($5M) Public Liability underwritten by Hiscox
- Contact: info@purelight.uk
- Privacy & Data Processing Policy: [api.purelight.uk/privacy](https://api.purelight.uk/privacy) (prompts never logged, never used for training, zero retention)

## The Edge Execution Chassis: Production Modules

1. **[Cleo SEO](https://cleoseo.com)** · Lead module
   - Category: Autonomous Edge SEO Operator
   - Core Tech: Rust HTMLRewriter stream interception (sub-5ms), strict WooCommerce dynamic-path bypass (cart/checkout/my-account/wp-admin/wc-ajax are never cached or transformed), 50ms fail-open transform deadline, Origin Shield snapshot failover.
   - Purpose: White-label edge SEO execution layer partnering with hosting platforms: automated structured data, Core Web Vitals, and page-builder bloat removal with zero origin PHP plugins.

2. **[AuditShield](https://auditshield.site)** · Compliance module
   - Category: EU AI Act & GDPR Continuous Compliance RegTech
   - Core Tech: Multi-standard regulatory audit graph, Article 14 human-oversight verification, automated evidence packages, Shopify zero-install App Bridge.
   - Purpose: Continuous regulatory audit and risk classification for enterprise systems deploying AI models across EU/UK jurisdictions.

3. **[PARLEY](https://parleyapp.site)** · AI agent governance layer (next layer of the chassis)
   - Category: Edge AI Agent Governance & Execution Firewall
   - Core Tech: Cloudflare Workers, Durable Objects, Cognitive Judo de-escalation, zero-leak streaming canary transforms.
   - Purpose: Low-latency CDN-edge mediation protecting origin APIs, databases, and servers from rogue autonomous AI agents, prompt injections, and multi-agent coordination swarms.

## Commercial Partnership Models
- **Wholesale White-Label Licensing:** Integration of Cleo SEO (lead module) directly into hosting control panels (cPanel, Site Tools), CMS marketplaces, and cloud platforms.
- **Enterprise Co-Branding & Distribution:** Co-developed edge compliance and security modules with revenue-share structures (e.g. 50/50 wholesale models).
- **Institutional Due Diligence & Investment:** Direct engagement with enterprise buyers, venture partners, and strategic corporate sponsors backed by full statutory transparency and £5M indemnity coverage.

## Verification & Key Facts
- Primary Domain: [purelight.uk](https://purelight.uk)
- Architecture: One unified edge execution chassis on a vendor-neutral edge mesh (Cloudflare Workers, D1, KV, Durable Objects, Azure AI Studio, Google Vertex AI).
- Institutional Inquiries: info@purelight.uk

## Moirai AI Model Family (in-house)

Pure Light trains and serves its own AI models at the edge under the Moirai family: Clotho (narrative & roleplay, anti-slop training), Lakhesis (editorial & content), Atropos (conversation & companion). First release: Clotho, OpenAI-compatible at [api.purelight.uk/v1](https://api.purelight.uk/v1) with 32,768-token context and published pricing of $0.10/$0.30 per million tokens. Quality is machine-validated against our open slop-detection methodology, with the A/B evidence published at [api.purelight.uk/slop-index](https://api.purelight.uk/slop-index) (Clotho 0.0 vs base 0.6). No prompt logging, no training on user data, zero retention.