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

jev-effort-router

Hermes Agent plugin where Jev picks both the model and reasoning effort for every turn on Ollama:Cloud.

Projectbeta
View on GitHub ↗

Overview

jev-effort-router makes model and reasoning-effort selection per turn for Hermes Agent. Hermes normally runs one model at one effort level for a whole session, so a two-line question and a hard refactor pay the same. This plugin's llm_request middleware asks Jev two Choice questions on every user turn (which of six benchmarked Ollama:Cloud models, and which effort level) and rewrites the outgoing provider request.

The author's point is that it decides both automatically per turn, where other routers pick only a model or leave effort to manual changes. A confidence guard (0.5 by default) protects against uncertain picks. Jev writes nothing; the selected model does all reasoning and generation. Scope is strict: it only routes the ollama-cloud provider, and every other provider passes through untouched. The README cites about 270 ms per decision at Jev's list price.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich Ollama:Cloud model should handle this turn?—
ChoiceWhich reasoning effort fits this turn?—

How it works

  1. User turn enters Hermes llm_request middleware
  2. Ask Jev model_route and reasoning_effort Choices
  3. Apply a confidence guard (default 0.5)
  4. Rewrite the provider request with the chosen model and effort

Strengths

  • Per-turn effort as well as model selection
  • Confidence guard against uncertain routes
  • Leaves other providers untouched

Caveats

  • Only works with Hermes on the ollama-cloud provider
  • Routing grid limited to six models

At a glance

Type
Project
Status
beta
Language
Python
Runtime
Hermes Agent plugin
Has code
yes

Use cases and integrations

How curated lists describe it

Frequently asked questions

What does Jev decide?

Which model and which reasoning effort to use for each turn.

Does it work with other providers?

No, it only routes ollama-cloud; other providers pass through unchanged.

What if Jev is unsure?

A confidence guard (default 0.5) prevents low-confidence routing.

Related projects

  • hermes-jev-router

    Hermes Agent plugin that picks the model each turn with a direct TypeSafe Jev Choice and enforces trim-then-compress context handling.

    PluginbetaPython
  • pi-jev-thinking-router

    Pi/omp extension that keeps the model fixed and has Jev choose only the thinking effort each turn via a local SPI daemon.

    LibrarybetaTypeScript
  • sabi

    Host-agnostic routing layer that picks model, effort and provider per round for coding harnesses from trajectory evidence.

    ProxybetaTypeScript
  • auto-model-router

    Cost-, cache- and quota-aware LLM router that classifies each turn with a Jev-class model, prices every route, and checks cheap answers.

    ProjectprototypePython
  • hermes-jev-router

    Experimental Hermes plugin returning routing plans: Jev rates task complexity, deterministic policy picks the cheapest eligible model.

    PluginprototypePython
  • HiRoute

    Local-first routing engine for long-running agent tasks that uses a Jev decider to pick a model group per turn.

    PluginproductionRust, TypeScript