llm-to-jev
Browser tool that finds bounded decisions in LLM prompts and proposes Jev Choice, Score and Noul questions with SDK code.
Overview
LLMtoJev looks at prompts written for GPT, Claude or Gemini and finds the bounded decisions inside them, proposing Jev Choice, Score and Noul questions. It is aimed at classification, scoring, routing and yes/no judgments that do not need generated prose. Each prompt is marked as fully, partially or not convertible, and for mixed prompts the tool separates Jev-compatible decisions from writing, summarisation, translation and other generative work that must stay with an LLM.
For a prompt like 'classify this ticket as billing, technical, account access or other, score its urgency from 0 to 1, and decide whether it needs human review', it proposes a choice, a score and a noul. It exports ready-to-review code for TypeSafe's official JavaScript and Python SDKs, plus Ruby and Go over HTTP, and a neutral JSON representation. Everything runs locally in the browser with no framework, server or account. The README stresses it is a conversion assistant, not a guarantee of equivalent behaviour. Tests and an optional live smoke test are included.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | Which parts of this LLM prompt are bounded decisions that Jev could answer? | fully convertible · partially convertible · not convertible |
How it works
- Paste an LLM prompt
- Compiler detects choice, score and yes/no decisions
- Prompt is rated fully, partially or not convertible
- Generative parts are left for the LLM
- Export SDK code or JSON for review
Strengths
- Helps identify migration candidates
- Runs fully in the browser
- Code export for several languages
Caveats
- Proposals need human review
- Heuristic detection may miss or mislabel decisions
- Does not guarantee equivalent behaviour
Getting started
git clone https://github.com/alexwestco/llm-to-jev.git && cd llm-to-jev && npm run dev, or use the live demo.At a glance
- Type
- Project
- Status
- beta
- Language
- JavaScript
- Runtime
- static browser app
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
Convert LLM prompts to Jev prompts.
hellogumbo/awesome-jev
Frequently asked questions
What does it produce?
Proposed Jev questions plus SDK code for JavaScript, Python, Ruby or Go.
Does it call Jev?
No, conversion runs locally; an optional live smoke test uses your key.
Will the converted version behave the same?
Not guaranteed; the README calls it a conversion assistant.
Related projects
DeepSeek Harness Web GUI plugin giving the agent a jev_ask tool, usage tracking and an optional command guard with explicit caveats.
PluginbetaTypeScriptEarly language design where typed Jev judgments are a runtime operator, with an engine that batches judgments over shared state.
ProjectprototypeTypeScriptSmall Next.js app to paste a state, build noul, choice and score questions, and inspect Jev's typed answers and distributions.
AppprototypeTypeScriptSkill that makes Claude Code and Codex split AI features into Jev, LLM or plain-code steps before writing code.
ProjectbetaTiny JavaScript semantic matcher that classifies text locally in microseconds and escalates ambiguous cases to Jev.
ProjectbetaJavaScriptSample Cloudflare Worker app with three playful demos, one per Jev primitive: Noul (is it a sandwich?), Score and Choice routing.
AppprototypeTypeScript