JevKit

Developer tool

Prompt → Jev

Turn an existing classification, routing, or scoring prompt into focused Choice, Score, and Noul decisions.

See which tasks are better handled by a generative LLM, deterministic code, or application logic.

Your prompt is sent through OpenRouter to an AI model provider to generate an AI-assisted draft. Do not paste API keys, credentials, secrets, or sensitive production data.

0 / 5000 characters

JevKit does not intentionally persist your raw prompt in an application database. OpenRouter and the routed model provider handle the request under their respective data policies.

Example input: Illustrative

How to use this draft

How it works

Paste an existing LLM prompt. The tool identifies bounded judgments, proposes JevKit draft questions, and marks tasks better left outside Jev.

What converts well?

Classification, routing, ordinal scoring, and a proposition that needs a probability are good candidates.

What stays elsewhere?

Keep summaries and free-text generation with an LLM. Keep exact counting, policy thresholds, and actions in code.

Choice, Score, or Noul?

Choice selects among named options. Score uses an ordered scale. Noul estimates the probability of a true/false proposition; it is not a fixed Boolean.

Review each generated instruction, criterion, and warning against your real workflow before using the draft. This is an AI-assisted JevKit format, not an official TypeSafe request payload.

This tool creates an AI-assisted JevKit draft for human review. It does not run Jev or produce an official TypeSafe request.

Primitive guidance: TypeSafe official documentation. Generation is routed through OpenRouter under its data policy and the model provider's policy. Sources checked 2026-09-24.