Jev demos
Six side-by-side demos where Jev makes the decision, Claude or Kimi writes the words, and code owns control flow.
Overview
Jev demos is a set of six live, side-by-side demos built around the pattern 'Jev makes the decision, an LLM writes the words, your code owns the control flow'. The LLM side can be switched between Claude (Haiku 4.5, Sonnet 5, Opus 5) and Kimi (K2.6, K3), and every tab, price and pipeline label follows the switch. A list summary mentions seven demos and PII handling among them.
It runs with no keys in a clearly badged simulated mode, and a presenter mode adds a narrator bar, spotlighting, guide cards, 'what to notice' callouts using real run numbers, and a two-minute guided tour. A list summary adds that each visitor's key gets its own budget by fingerprint and keys are redacted from upstream errors. It requires Node 22.6+ and runs TypeScript natively without a build step.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which model tier should write the response?(inferred) | — |
| Yes / no | Does this text contain PII?(inferred) | — |
How it works
- Pick a demo tab
- Jev answers typed questions in a few hundred milliseconds
- Code routes to the selected LLM only where text is needed
- Show timing, cost and a narrated explanation
Strengths
- Runs with no keys in simulated mode
- Presenter mode with guided tour
- Switchable LLM side for cost comparison
Caveats
- Demo focused; simulated outputs are not real decisions
- Specific demos not detailed in the excerpt
Getting started
`git clone git@github.com:mayank953/Jev.git && cd Jev && npm install && npm start`, then open http://localhost:3000.At a glance
- Type
- App
- Status
- prototype
- Language
- TypeScript
- Runtime
- Node 22.6+ web app
- Has code
- yes
Use cases and integrations
How curated lists describe it
Seven side-by-side demos that run with no keys and label themselves simulated. Each visitor's key gets its own budget by fingerprint, and any key is redacted out of upstream errors.
MrJev/awesome-jev
Frequently asked questions
What is the pattern shown?
Jev decides, an LLM writes, and code controls the flow.
Can I try it without keys?
Yes, missing keys trigger a clearly labelled simulated mode.
Which LLMs are supported?
Claude (Haiku, Sonnet, Opus) and Kimi (K2.6, K3).
Related projects
Korean course code for attaching Jev to Claude Code, Codex and OpenClaw: skills, hooks, routers and open-model comparisons.
ProjectproductionJavaScriptCompanion code for the ZazenCodes course on decision models: primitives notebook, use-case examples and a Pi model-router build.
ProjectproductionPythonFree 12-notebook course building 13 agent tools with Jev as the fast decider and an LLM as the slow writer, for under $0.20.
ProjectproductionPythonRouter that uses Jev's cloud decisions today, logs every decision with probabilities, and distils a local routing model from your traffic.
ProjectbetaPythonBeginner manual, video script and three small examples for building an agent harness with Jev via langchain-typesafe.
ProjectprototypePythonGuide and small prototype for using Jev with Herdr to route tasks to coding agents like Claude, Codex and Hermes under code-owned policy.
Projectprototype