hermes-jev-helper
Hermes Agent plugin that classifies each turn's route (skill, research, history, others) with Jev before the model call.
Overview
hermes-jev-helper, documented in Spanish, externalises an agent harness's routing decision to a typed intent classifier that runs before the model call and appends natural-language routing guidance to the turn. It does not execute the task or produce the answer; it picks one of four routes (skill, research, history or others) and states a confidence. The goal is not better answers but less route improvisation, so the same kind of request gets the same kind of route.
It is a Hermes Agent pre_llm_call plugin, with the reasoning for that choice documented in a case study. It calls Jev through OpenRouter's decisions endpoint by default, with configurable endpoint, model slug and credential variable resolved through the harness's secret scopes, environment or Infisical helper. Without a credential, it does nothing and never breaks the turn. Settings can be edited in the desktop app.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which route should this turn take? | skill · research · history · others |
How it works
- Before the model call, send the turn to Jev
- Jev picks a route with confidence
- Plugin appends routing guidance to the turn
- No credential means no action
Strengths
- Consistent routing for similar requests
- Never breaks the turn
- Configurable from the desktop app
Caveats
- Spanish documentation
- Guidance only; the model may still deviate
Getting started
hermes plugins install maurorosero/hermes-jev-helper, hermes plugins enable hermes-jev-helper, and provide OPENROUTER_API_KEY.At a glance
- Type
- Plugin
- Status
- beta
- Language
- Python
- Runtime
- Hermes Agent plugin
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Hermes harness plugin — Jev-backed intent classifier that routes each turn to a…
wh000wh000/awesome-jev-live
Frequently asked questions
What routes exist?
skill, research, history and others.
Does it answer the user?
No, it only adds routing guidance before the model call.
What if no key is set?
It does nothing and does not fail.
Related projects
Hermes Desktop plugin where Jev maps misspelled, shortened or meaning-based slash commands to real ones at ≥0.85 confidence.
PluginbetaJavaScriptPre-flight hook for WorkBuddy and Hermes Agent where Jev classifies each prompt on nine dimensions and injects routing cues.
ProjectbetaPythonAsync Python intent and tool router: Jev picks the handler, and a code-side confidence gate stops weak or invented decisions.
ProjectbetaPythonOpt-in Hermes Agent plugin that asks Jev to name the one installed skill worth loading before the model call.
Agent skillbetaPythonConfidence-aware skill router that uses Jev to pick agent skills from large catalogues, declining or asking for review on weak matches.
Agent skillresearchPythonPage companion where a site's actions are WebMCP tools and Jev picks the action a visitor's request means, asking back when unsure.
Libraryprototype