Understand the system, not the contents of your request.

To perform a decision, jevify sends the submitted state, instructions, criteria, and questions to the selected decision provider. To perform a repository audit, jevify reads the repository URL, supplied files, or uploaded content needed to produce the requested report.

This content is processed to provide the requested result. jevify does not currently run a second language model to classify, redact, summarize, or vectorize customer prompts after the request. It does not currently generate semantic taxonomies or embeddings from customer prompts for analytics.

For reliability, quotas, abuse prevention, and aggregate product analysis, jevify records:

• request time, origin, completion status, latency, and safe error code;
• provider and model used, including provider attempts and fallbacks;
• input and output token counts and known cost, when supplied by the provider;
• number of items, questions, and decision dimensions;
• question types used—Noul, Choice, or Score—and the character count of the submitted state;
• the associated account and API key record, when applicable.

This event record does not include the submitted state, prompt text, instructions, criteria text, provider response, decision answers, repository contents, or full API key.

Sign-in providers supply the account identity needed to authenticate you. jevify stores account-linked API key names, prefixes, hashes, creation and usage dates, and revocation status. A generated jvf_… key is shown once; only its cryptographic hash is retained.

If you save a TypeSafe API key, it is encrypted before storage, used only on the server to make requests you initiate, and never displayed again in full. You can remove it from the account page.

Depending on your selected mode and the feature you use, request content may be transmitted to TypeSafe AI or the Lovable AI gateway for decisions, and to GitHub when you request analysis of a repository. Authentication is handled by your selected sign-in provider. Firecrawl is used to discover public case material. Their processing is governed by their own terms and privacy policies.

The public homepage decision demonstration processes only its fixed demonstration messages through an available provider, which may include OpenRouter, Vercel AI Gateway, or the Lovable AI gateway. It does not submit visitor-entered customer content.

You may revoke jevify API keys and remove a saved TypeSafe credential from your account. To request access to or deletion of account-linked data, email privacy@nxlv.ai from the email address associated with your account. Some minimal records may be retained where required for security, legal obligations, or dispute prevention.