Camazotz Consulting
An AI-automation consulting site whose hero demo is a real LLM call, not a video. The site is statically exported from Next.js to Cloudflare Pages; the document-summarization playground runs as a Cloudflare Pages Function calling Workers AI directly, with per-IP daily rate limiting backed by Workers KV.
The site presents an AI automation consulting practice focused on helping teams integrate AI-assisted development workflows. The centerpiece is a live document-summarization demo on the hero — a real Workers AI inference call, not a video or animation — that lets visitors see the service in action without booking a consultation.
The architecture is designed for zero operational overhead. The site is statically exported from Next.js and deployed to Cloudflare Pages. Server-side logic — just the AI demo handler and a contact form — runs as Cloudflare Pages Functions on the edge runtime, with no traditional origin server. The local development command runs the same static artifact that deploys to production, eliminating environment drift.
The AI demo calls Qwen3-30B on Workers AI with structured JSON output, a system prompt exemplar to pin response format, and chain-of-thought suppression for lower latency. Rate limiting uses Workers KV with per-IP daily caps that also cover preview deployments, preventing bypass via branch URLs.
The front-end shows three animated progress stages while the LLM call runs. Since the models typically respond within a couple seconds, the stages are scheduled optimistically before the fetch resolves and cancelled on response — creating a polished experience without masking the fact that the inference is live.