Build Jev Schemas From Plain Language
Transform decision descriptions into valid Jev JSON for routing, classification, and guardrails. Free, no signup.
Your Jev Schema Will Appear Here
Describe a decision and click Generate
Five free generations an hour. Help keep it that way.
These tools use real AI credits. Reader support keeps them generous, private, and available without ads, accounts, or a paywall.
What support protects
5 generations per hour
No ads or paywall
The best available model
No tracking of your writing
What happens below the $100 hosting floor?
Public use may be reduced to 3 generations per day, and a lower-cost model introduced after Sep 30, 2026.
September hosting floor
$47 of $100
Keeps the tool online, fast, and free for everyone.
Raises the public cap to 10 generations per hour.
Higher-quality AI models for better, more coherent outputs.
Time to fix bugs and ship the features you ask for.
Advanced features, tool improvements, and your feature requests.
If these tools have been useful, a monthly pledge or one-time coffee genuinely helps.
Secure payment · no account for one-time support
AI Outfit Generator
Generate complete outfit ideas with AI. Pick an occasion, style, and season — or paste your closet to build looks from what you own. Free, no signup.
Instagram Bio Generator
Create niche-specific Instagram bios under 150 characters with emojis, line breaks, and CTAs. AI-powered, free, no sign-up.
What is a Jev Question Builder?
The Jev Question Builder transforms plain-language decision descriptions into valid Jev JSON schemas. Jev is a decision model API that returns typed, probability-backed answers instead of generated text — ideal for routing, classification, moderation, and guardrails in production applications.
Describe what you want to decide in natural language, and the tool generates a complete Jev request body with properly structured instructions and criteria. Supports all three Jev question types: Choice (pick one option), Score (rate on a scale), and Noul (yes/no judgment).
Key Features
Plain Language to Structured Schema
Describe your decision in everyday language — 'route support tickets to the right team' or 'rate how urgent this message is' — and get a properly formatted Jev question with clear instructions and well-defined criteria.
All Three Jev Question Types
Generates Choice questions for classification (up to 255 options), Score questions for ordered ratings (2–10 levels), and Noul questions for yes/no judgments. Let the AI pick the best type or specify it yourself.
Best Practices Built In
The generated schema follows Jev's documented best practices: distinct descriptions for every option, 'other' catch-all for Choice questions, proper low-to-high ordering for Score levels, and affirmative phrasing for Noul.
Ready-to-Use JSON Output
Copy the complete JSON request body directly into your code. Includes the model version, sample state, and properly structured questions map — ready for the Jev API via OpenRouter.
Free for All Developers
No sign-up, no API key required. Generate as many Jev schemas as you need while building and testing your decision workflows.
Perfect for everyone
- Developers integrating Jev into applications
- Engineers building AI-powered routing and classification
- Teams implementing content moderation and guardrails
- Anyone learning the Jev API
How to Use the Jev Question Builder
Transform plain-language decisions into valid Jev JSON schemas in three simple steps
Describe Your Decision
Enter a plain-language description of the decision you want Jev to make. Be specific about the options, categories, or judgment criteria. For example: 'Route customer support tickets to shipping, refunds, billing, or general support based on the message content.'
- Mention specific options or categories if you have them in mind
- Describe what input (state) the decision will be made on
- Include any edge cases or 'other' scenarios
Choose Question Type (Optional)
Let the AI automatically choose the best question type, or specify it yourself. Use Choice for classification into categories, Score for ordered ratings or severity levels, and Noul for simple yes/no judgments.
Generate and Copy JSON
Click generate to create a complete Jev request body. The output includes the model version, sample state, and properly structured questions with clear instructions and criteria. Copy it directly into your code.
Send to Jev via OpenRouter
Use the generated JSON as the request body for POST requests to OpenRouter's API with model 'typesafe/jev-1.13'. The response includes typed answers with probabilities you can use directly in your application logic.
Who Uses the Jev Question Builder?
Build Jev decision schemas for routing, classification, moderation, and more
Support Ticket Routing
Automatically route incoming support tickets to the right team. Generate Choice questions that classify messages into shipping, billing, refunds, technical support, or other categories based on the ticket content.
- Route tickets to specialized teams instantly
- Include an 'other' catch-all for edge cases
- Get probability-backed confidence for each routing decision
Survey Response Coding
Analyze open-ended survey responses at scale. Create Noul questions for each theme in your codebook to tag responses with multiple categories, then aggregate the results for reporting.
- Code thousands of responses in minutes
- Multi-tag with one yes/no question per theme
- Validate against hand-coded samples
Content Moderation and Guardrails
Build moderation pipelines that flag content based on specific criteria. Use Score questions for severity ratings or Noul questions for policy violation checks, with thresholds you control.
- Set confidence thresholds for automated actions
- Combine multiple checks in a single request
- Keep humans in the loop for borderline cases
Intent Classification
Classify user intents in chatbots and voice assistants. Generate Choice questions that map user messages to predefined intents, with probabilities that help you handle ambiguous inputs.
- Map messages to intents with confidence scores
- Handle 'none of the above' with an other option
- Use probabilities to trigger clarification flows
Tips for Best Results
Write better Jev questions with these best practices from the official documentation
State the Exact Condition
Jev answers the question you wrote, not the one you meant. Be specific in your instructions. When you look at a wrong answer and find yourself explaining what you really meant, that explanation is the missing half of the instruction.
Keep Arithmetic in Code
Jev is not a calculator. Don't ask it to count items, sum values, or compare dates. Extract the semantic judgment with Jev, then do the math in your application code.
Use Distinct Option Descriptions
Every Choice option needs a clear, unique description that tells Jev when to select it. Overlapping or vague descriptions lead to inconsistent results. Include boundary cases in the criteria.
Phrase Noul for Affirmative = High
Write Noul instructions so that a high probability (near 1) means 'yes' and low (near 0) means 'no'. Inverted phrasing confuses the model and hurts accuracy.
Send Only Relevant State
Large states full of irrelevant detail act as distractors. Filter your input to include only what the decision needs. You also pay per input token, so trimming saves money.
Test with Reordered Options
Jev can lean toward the first option in a Choice. Reorder your options and verify the answer is consistent. If it changes, your descriptions need more clarity.