Privacy & operating limits

Manual processing

Manual conversion, normalization, scheduling and diagram rendering run in your browser. This application has no account database, saved input history, session replay or analytics SDK. Downloads are initiated by you. Hosting providers still process ordinary request metadata under their own policies.

When you choose Ask Jev

After explicit consent, the context described beside that button is sent to our Cloudflare Worker and TypeSafe. This can include sample values, source text, definitions, availability or questions. Do not submit confidential or personal information without authority. The application does not persist these inputs or log provider bodies. TypeSafe processing and retention are governed by its terms, not a claim that inference is fully local.

Public-page reading

Read public page sends the URL to our server, which fetches only approved HTTPS documentation hosts and checks every redirect. It forwards no browser cookies, executes no scripts, accepts only HTML/plain text, and caps redirects, bytes and elapsed time. Unsupported sites can be used by pasting text. This is not an authenticated browser or paywall bypass.

Secrets and spending

TYPESAFE_API_KEY is a server-side Worker secret. The browser never receives it. Requests are limited to 16 KB and a bounded set of server-authored questions; inference has a 15-second timeout and no automatic retries. Client and aggregate limits are 6 and 30 requests per minute per Cloudflare location. These limits are eventually consistent and are not a strict global billing cap. Provider-side spend limits are required before public AI launch.

Output boundaries

Typed choices and confidence do not guarantee correctness. Review proposals. Entity extraction is capped and may omit mentions. Source excerpts are not independent fact verification. Missing credentials, protection or valid output produce an explicit failure, not fake model results.