Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

Jev Auto Router

Prototype per-call model router for Codex where Jev picks the GPT model and reasoning effort for each call via a local Responses proxy.

ProxyprototypeApache-2.0
View on GitHub ↗

Overview

Jev Auto Router is a Chinese-documented prototype (with an English README) that re-selects the model and reasoning effort on every meaningful model call within a Codex session. A local Responses proxy keeps the Codex session and tool loop continuous; for each call Jev makes one typed Choice over the (model, effort) pairs the host can actually request, currently three GPT-6 tiers, and the proxy applies only the model and reasoning-effort change. Failures, shadow mode, privacy refusals or insufficient information fall back to a fixed baseline.

After the task, independent verification checks whether it still passed, and Router Compass brings routing choices, actual usage and verification results together to answer whether using the flagship less still completes tasks correctly and reduces cost. The README is explicit that this is a validation prototype: the runtime passes deterministic HTTP tests, but real-machine evidence and paired-evaluation conclusions are still pending, there is no production install, and it makes no cost-savings claims. Jev only sees allow-listed facts and never upstream credentials. It is Apache-2.0 licensed.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich (model, reasoning effort) pair should handle this call?gpt-6-luna · gpt-6-sol · gpt-6-astra (gated)

How it works

  1. Codex sends a Responses request to the local virtual model
  2. Proxy builds allow-listed routing state and candidate pairs
  3. Jev makes one Choice; guard checks it
  4. Proxy rewrites only model and effort and forwards
  5. Verify the task afterwards and record in Router Compass

Strengths

  • Per-call granularity rather than per-task
  • Jev never sees credentials
  • Explicit fallback baseline and shadow mode

Caveats

  • Prototype without real-machine evidence yet
  • No production install path
  • No cost-savings claims

At a glance

Type
Proxy
Status
prototype
Language
TypeScript
Runtime
local Responses proxy (Node 22+)
Has code
yes
Has tests
yes
License
Apache-2.0

Use cases and integrations

How curated lists describe it

  • Coding agents: per-call Codex GPT routing where Jev makes one typed Choice over host-available (model, effort) pairs; a local Responses proxy keeps the tool loop continuous, then independent verification and Router Compass record whether the task still passed (prototype).

    yibie/awesome-jev

  • Unofficial model-routing prototype that uses Jev to choose a model and reasoning effort for each call, with a local Responses proxy and independent task verification

    AnotiaWang/awesome-decision-models

  • Jev Auto Router (Jev Router): experimental per-call GPT model routing for Codex via TypeSafe Jev and a local Responses proxy, with independent task verification.

    logicrw/awesome-jev-projects

  • Prototype per-call router for Codex: a Jev Choice picks the model and reasoning effort for each call, a local Responses proxy continues the tool loop, and an independent verification step audits the result; no cost-savings claims.

    cobanov/awesome-jev

  • Per-call GPT model routing for Codex: Jev chooses model and effort, a local Responses proxy keeps the tool loop intact, then independent verification and Router Compass check whether the task still passed.

    hellogumbo/awesome-jev

Frequently asked questions

What does Jev decide?

Which GPT model and reasoning effort each Codex call should use.

Does it save money?

Unproven; the README makes no savings claims and evaluation is pending.

Is it ready to install?

No, it is a validation prototype without a production install flow.

Related projects

  • jev-router★ 593

    Per-turn model routing for Claude Code and Codex: wraps the real CLI and lets Jev send simple turns to fast models and hard ones to strong.

    ProxybetaJavaScript
  • jev-codex-router★ 274

    Per-call router for Codex where Jev picks the model and thinking effort for every model call, including tool continuations.

    ProjectbetaPython
  • coding-router-jev

    TypeScript/Bun launcher that asks Jev which model and reasoning effort each Codex, Claude Code or Pi turn should use.

    ProjectbetaTypeScript
  • frost

    Go CLI model router that uses one Jev analysis to recommend a model, harness and effort level, then selects deterministically.

    CLIbetaGo
  • jev-codex-bridge

    Local bridge where Jev selects model and reasoning effort for Codex Desktop and CLI, with a Windows service and cache-aware routing.

    CLIbetaTypeScript
  • jev-codex-model-and-effort-router

    macOS wrapper for Codex that lets Jev pick the model and reasoning effort for a thread, then pins that route.

    Projectbeta