jev-use-cases
Video demo app showing Jev use cases with the AI SDK and shadcn/ui, from intent search and model picking to chat moderation.
Overview
Jev use cases is a Next.js demo app built for a video about Jev. Each page corresponds to one segment of the script: intent search, small typing interactions, a model picker, response depth, json-render and Twitch chat moderation. Jev answers Choice, Score and Noul questions through the AI SDK's experimental_evaluate, and app code decides what the UI does with the probabilities: apply when confident, suggest when unsure, do nothing when low.
Questions per demo are built on the server, where the key stays. A confidence-tier module and a calm UI hook add hysteresis so the interface does not flicker between states. Keys can be a direct TypeSafe key, OpenRouter (which also streams an LLM answer in the response-depth demo) or Vercel AI Gateway; with no key every demo falls back to keyword rules so the app still works offline. The header shows which source answered and the real round-trip time, and a switch reveals the probabilities.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which model should handle this request? (model picker demo)(inferred) | — |
| Score | How deep should the response be?(inferred) | — |
| Yes / no | Should this chat message be moderated?(inferred) | — |
How it works
- User interacts with a demo page
- Server builds questions and calls experimental_evaluate
- Confidence tiers decide apply, suggest or nothing
- UI updates with hysteresis; probabilities optionally shown
Strengths
- Several UI patterns in one app
- Offline keyword fallback
- Key never leaves the server
Caveats
- Demo for a video, not a library
- No evaluation of accuracy
Getting started
bun install, cp .env.example .env.local (add a key or skip), bun run dev.At a glance
- Type
- Library
- Status
- prototype
- Language
- TypeScript
- Runtime
- Next.js
- Has code
- yes
Use cases and integrations
How curated lists describe it
Demo app: Jev (TypeSafe System One) use cases with the AI SDK and shadcn/ui.
hellogumbo/awesome-jev
Frequently asked questions
What demos are included?
Intent search, typing interactions, a model picker, response depth, json-render and Twitch chat moderation.
Does it work without a key?
Yes, every demo falls back to keyword rules offline.
How are low-confidence answers handled?
Confidence tiers decide whether to apply, suggest or do nothing.
Related projects
React components that use Jev judgments to choose which component renders, how lists are ordered and whether affordances show.
ProjectbetaTypeScriptVercel Labs' generative UI framework with an experimental compose path where Jev picks which catalog components and actions to render.
ServicebetaTypeScriptCommunity shadcn-style React components and interactive lab for building TypeSafe AI interfaces.
ProjectbetaTypeScriptPlayground comparing generative UI via an LLM writing json-render specs with Jev choosing and arranging prepared components.
LibraryprototypeTypeScriptLibrary that composes typed Jev judgments like reactive state in React and backend programs, with automatic batching and reuse.
HookbetaTypeScriptOne-file Mastra input processor that asks Jev whether a message must be blocked, failing open behind a deadline and circuit breaker.
ProjectproductionTypeScript