How Tinplate handles requests, zero-logging isolation, and optional edge caching.
Tinplate APIs inspect and process request payloads strictly in volatile V8 isolate memory on Cloudflare Workers for the duration of the HTTP request and immediately discard them. We never log, store, sell, or retain your request bodies, response bodies, signing secrets, PII, HTML, emails, or tokens.
Cloudflare Workers Observability, Invocation Logs, and Analytics Engine bindings are explicitly disabled
(observability.enabled = false, head_sampling_rate = 0, invocation_logs = false)
across both the Gateway and Fetcher Workers:
Security-minded buyers should account for how requests reach Tinplate's Cloudflare Workers:
*.rapidapi.com) or Zyla API Hub, your HTTP request traverses that marketplace's API proxy infrastructure before reaching Tinplate's Cloudflare Worker. While Tinplate never logs or stores payloads, the marketplace proxy processes your request under RapidAPI's Privacy Policy or Zyla's Privacy Policy.Authorization: Bearer <API_KEY>) directly against https://api.tinplate.dev. Direct requests terminate straight at Tinplate's Cloudflare edge isolate with zero marketplace proxies in the path./v1/eu-vat-iban/vat) and public HTML-to-Markdown conversions by URL hash (/v1/html-to-markdown) may be cached briefly in Cloudflare KV to reduce upstream latency. Raw HTML inputs and private secrets are never cached."use_ai_judge": false), /v1/prompt-guard and /v1/hallucination-diff run 100% deterministically inside the Worker isolate with zero external calls. When you explicitly pass "use_ai_judge": true, the text excerpt is passed over an internal Cloudflare binding to Cloudflare Workers AI (@cf/meta/llama-3.1-8b-instruct) for stateless inference; neither Tinplate nor Cloudflare trains models on or retains Workers AI inputs/outputs.