Short on purpose
Privacy
This website uses privacy-friendly, cookieless Vercel Web Analytics to measure aggregate pageviews without storing personal data or tracking across sites. It is static files served from Vercel with a GitHub Pages mirror. The only private data involved is what you choose to send to the Jev endpoint when you run the CLI yourself.
{ "site": "cookieless aggregate metrics",
"cli": "rows sent only to api.typesafe.ai",
"scope": "per request, fan-out batches",
"storage": "clean.jsonl / rejected.jsonl, local" }
Website
Static pages plus two scripts: the menu and copy helper, and Vercel Web Analytics, which counts aggregate pageviews without cookies or cross-site tracking. No personal data is stored. The only third party is the hosting provider's standard serving logs.
Running the CLI
The CLI sends each row to the TypeSafe endpoint as request state, wrapped in fan-out batches. What leaves your machine is the row content plus the preset's questions. The answers come back as typed verdicts; accepted rows land in clean.jsonl and rejected ones in rejected.jsonl, both written locally. Pre-filtering drops blank and padding lines before anything is sent.
API keys are read from the TYPESAFE_API_KEY environment variable or the CLI argument, sent as the bearer token to the endpoint, and never logged by the tool. The dry run mode performs the whole pipeline against a local mock server and sends nothing to the network.
You hold the volume knob
Whether a row is sensitive is your call. The tool offers no server-side storage: there is no account, no cloud upload except the endpoint request itself, and no telemetry. If the dataset is private, the same caveat that applies to any API-backed pipeline applies here: the row content reaches the model provider.