ask-jev-ai
Public wall where Jev answers short questions yes, no or it depends in about 100 ms and picks which LLM answers signed-in users.
Overview
Ask Jev is a live public app at askjev.ai. Anyone can ask a question of three to fifteen words, and Jev answers yes, no or it depends in about 100 milliseconds; every judged ask appears on a realtime wall with a running count toward one million and the exact cost. Signed-in users can ask up to sixty words, Jev picks which model should answer (Gemini Flash Lite, Claude Haiku, GPT mini or Perplexity Sonar), and a short answer streams in through the Convex AI Gateway with the model's name and a one-line reason.
Each ask goes to Jev with six questions: the yes/no/depends verdict, whether it is unkind, adult or targets a person, its mood and its topic. Cards show verdict and confidence, flag close calls under 60%, and can reveal the runner-up. The whole app is open source, built from Convex components (Auth, Agent, Sharded Counter, Rate Limiter, Static Hosting) and the Convex AI Gateway, so it stores no provider keys.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Does Jev say yes, no, or it depends? | yes · no · it depends |
| Yes / no | Is it unkind / adult / does it target a person? | yes · no |
| Choice | What is the mood / topic? | — |
| Choice | Which model should answer this ask? | Gemini 3.5 Flash Lite · Claude Haiku 4.5 · GPT 5.4 mini · Perplexity Sonar |
How it works
- User submits a short ask
- Convex action sends it to Jev via the AI Gateway with six questions
- Moderation answers decide whether it fits the wall
- Verdict, confidence, mood, topic, cost and latency show on the card
- For signed-in users, Jev picks a model to stream a short answer
Strengths
- About 100 ms verdicts per the README
- Moderation and routing from one call
- No provider keys stored in the app
Caveats
- Built on alpha and beta Convex components
- Public wall depends on moderation thresholds
Getting started
Clone the repo and run it against your own Convex deployment (backend in convex/, app in src/).At a glance
- Type
- App
- Status
- production
- Language
- TypeScript
- Runtime
- Convex backend + React app
- Has code
- yes
Use cases and integrations
How curated lists describe it
Most AI demos generate text. Jev does not. It reads a sentence and returns typed answers with probabilities: a choice, a yes or no, a score. That makes it usable as a primitive inside ordinary code rather than a chatbot bolted onto a page.
logicrw/awesome-jev-projectsA public wall where anyone asks a question in three to fifteen words and Jev, TypeSafe's judgment model, answers yes, no, or it depends in about 100 milliseconds. Every judged ask lands on the wall in realtime, with a running count toward one million, showing cost.
hellogumbo/awesome-jev
Frequently asked questions
Does Jev write the answers?
No, Jev gives the verdict and picks a model; the chosen LLM writes any text answer.
How is the wall moderated?
Jev also judges whether an ask is unkind, adult or targets a person.
Where does the key live?
Calls go through the Convex AI Gateway on a deployment token, so the app stores no provider keys.
Related projects
Korean course code for attaching Jev to Claude Code, Codex and OpenClaw: skills, hooks, routers and open-model comparisons.
ProjectproductionJavaScriptPython CLI that shows which OpenAI model Jev would route a query to, independently reviewed by a local Ollama model.
ProjectprototypePythonStudy finding commercial decision models including Jev give different yes/no probabilities on Taiwan statements by Chinese script.
ServiceresearchPythonLangChain's Python framework, which includes an optional TypeSafeClassifier runnable plus model-routing and risky-tool middleware using Jev.
PluginproductionPythonLangChain.js, which includes an optional TypeSafeClassifier runnable and middleware for bounded routing and tool-call checks with Jev.
PluginproductionTypeScriptWordPress plugin connecting sites to Jev, with PHP helpers for choice, score and yes/no decisions plus a REST proxy and JS client.
ProjectbetaPHP