User repository on GitHub
Route HTTP requests by meaning. A semantic router for Hono powered by Jev.
- Forks
- 4
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Repos & SDKs
Hono Jev Router
Experimental Hono router that matches an incoming HTTP request to a plain-language route description: one yes/no Jev question per description is evaluated in a single call, and the first registered route above the threshold wins.
- Stars
- 52
- Language
- TypeScript
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- yusukebe
- Last verified
- Sep 2026
About this project
hono-jev-router is an experimental router for Hono. Instead of a method and a path, you describe a request in plain words, and Jev by TypeSafe AI decides which description fits the incoming request. Do not use it for authentication or authorization. The request is the input of the model, so whoever sends the request can try to steer the answer. 'suspicious automated traffic' is a heuristic, not a security boundary.
Quoted from the project's README. Read the full README.
Highlights
- Registers semantic routes with app.on('jev', '<description>', handler) alongside normal Hono paths.
- One yes/no Jev question per description is evaluated in parallel in a single call.
- The first registered route above the threshold wins; unmatched requests fall through to notFound().
- Authorization, Cookie, and API-key header values are redacted before the request goes to Jev.
- Can run on Cloudflare Workers AI through the AI binding instead of an API key.
Quickstart
ts
import { Hono } from 'hono'
import { JevRouter } from 'hono-jev-router'
const app = new Hono({
router: new JevRouter({ apiKey: process.env.TYPESAFE_API_KEY }),
})
app.on('jev', 'a request from an AI agent', (c) => {
return c.text('# Documentation', 200, { 'Content-Type': 'text/markdown' })
})
export default appWatch out
MIT-licensed and explicitly experimental. Every semantically routed request is a paid model call that adds latency, the answer is a probability rather than a guarantee, and it must not be used for authentication or authorization.
Questions about this entry
Short answers drawn from the linked project or page.
No, and the README says so plainly. It warns explicitly not to use it for authentication or authorization, because the request is the model input, so whoever sends it can try to steer the answer. The suspicious automated traffic description is a heuristic, not a security boundary. Keep a normal rate limiter in front of it.
The method, URL, headers and the first maxBodyLength bytes of a textual body, default 4096. Values of Authorization, Cookie, Proxy-Authorization, X-API-Key and X-Auth-Token are replaced with a redaction marker, and that header list is configurable through redactHeaders. Everything beyond those bytes still leaves your server.
Every semantically routed request is one model call you pay for plus the added latency, and the answer is a probability rather than a guarantee. A description matches once its probability reaches the threshold, default 0.5, and the first registered match wins because order matters. Requests matching nothing fall through, usually to notFound().
Yes, through the AI binding. Jev is available as typesafe/jev on Workers AI, so you pass run and call the AI binding with the gateway option instead of an apiKey, casting c.env.AI if the model is not in your generated types yet. The choose option also lets you replace the decision entirely, which is handy for tests, caching or delegating to something that holds the credentials.
npm i hono-jev-router, then hand the router to Hono as `new JevRouter({ apiKey })` and register semantic routes with `app.on('jev', '<description>', handler)` next to ordinary path routes. It is MIT-licensed and published by Yusuke Wada, and the README calls it an experiment outright.
The decision is available as `c.get('jev')`, typed as JevResult: route, confidence, and a probabilities map keyed by every registered description. The README is explicit that those probabilities are independent of each other, so they do not sum to 1 and should be read as independent confidences rather than a distribution.
Yes. Path routes and middleware are handled by Hono's own TrieRouter, and a path route that returns a response wins before Jev is asked at all, so `app.get('/health', ...)` never reaches the model. Within the semantic routes, order decides: put specific descriptions first and broad ones last, because the first registered match wins.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Facebook: Intake Router keeps diagnosis out of the answer space at the front desk.
FacebookClinicnotes
X: OpenRouter Decision Model Rankings
X: herdr + jev, an agent router
X: A group chat bot with decision room routing
More Jev resources like this
A Juspay TypeScript SDK that treats Jev as a third inference type beside generate and stream: providers declare which kinds they serve, and tryDecide returns typed boolean, choice, or score answers, null when no decision provider is set.
Jev Repos & SDKs#community#typescript#sdk
Community
juspay/neurolinkTypeScript
juspay/neurolinkTypeScriptOpen multilingual System 1 decision models with published checkpoints for choice, score and noul questions, plus a router that dispatches each request to the right checkpoint in one forward pass.
Jev Repos & SDKs#community#python#open-models
Community
NandhaKishorM/layaPython
NandhaKishorM/layaPythonA filesystem-first framework for durable AI agents whose auto model router defaults to Jev through Vercel AI Gateway and whose evaluate helper asks typed Choice, Score, and Boolean questions inside tools.
Jev Repos & SDKs#community#typescript#agents
Community
vercel/eveTypeScript
vercel/eveTypeScriptAn ACP and MCP adapter that bridges TypeSafe Jev with any LLM: native Jev when TYPESAFE_API_KEY is set, or Codex, Claude, Grok, OpenCode, and OpenAI-compatible endpoints standing in as System One when it is not.
Jev Repos & SDKs#community#typescript#mcp
Community
tacticocc/JevbridgeTypeScript
tacticocc/JevbridgeTypeScriptAn OpenTelemetry exporter wrapper that scores each log record with Jev for diagnostic value, priority, and routing, letting low-value events skip an LLM-analysis branch while every record stays in your archive.
Jev Repos & SDKs#community#typescript#routing
Community
reachjalil/jevlogsJavaScript
reachjalil/jevlogsJavaScriptTyped semantic decisions for TypeScript: ask Zod 4 or native question batches about your data and get validated answers through async/await or Web Streams, with confidence gates, retries, replay and observability. Jev runs via TypeSafe AI or Vercel evaluation.
Jev Repos & SDKs#community#catalog-import#logicrw
Community
nitoba/questionsTypeScript
nitoba/questionsTypeScriptFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.