hermes-jev
Nerve: asynchronous System-1 supervisory plugin for Hermes Agent judging relevance, completion and recovery with Jev, Laya or OpenJev.
Overview
Nerve (hermes-jev) is an asynchronous System-1 supervisory layer for Hermes Agent. It acts as a watchdog and forecaster, covering relevance, completion, recovery and optional admission decisions, while the main Hermes orchestrator and reviewer keep final stop or continue authority. The core plugin is dependency-free and model runtimes run in sidecars.
Version 0.3.1 is the stable maintenance release of the profile-aware 0.3.0 line, focusing on fixes, hardening, completion and retry integrity, configuration correctness and an opt-in ask-only Action Gate, while preserving earlier behaviour when no profile is selected. Development builds add a provider-neutral backend seam: Jev remains authoritative by default, a shadow mode logs paired Laya or OpenJev decisions, and a crash-safe Jev, Laya and OpenJev matrix runner supports open-backend validation. Hermes catalogue installs are pinned to reviewed commits.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is the current step relevant to the goal?(inferred) | yes · no |
| Yes / no | Is the task complete?(inferred) | yes · no |
| Choice | Which recovery action fits this failure?(inferred) | — |
How it works
- Observe Hermes Agent activity asynchronously
- Ask the decision backend about relevance, completion and recovery
- Forecast and flag; optionally gate actions with ask-only
- Hermes orchestrator makes the final call
Strengths
- Does not block the main agent
- Open-backend shadow comparisons
- Pinned, reviewed releases
Caveats
- Development line differs from stable release
- Complex configuration
Getting started
Install from the Hermes plugin catalogue (pinned SHA) and configure a Jev key or Laya/OpenJev sidecar.At a glance
- Type
- Project
- Status
- beta
- Language
- Python
- Runtime
- Hermes Agent plugin + sidecars
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
An asynchronous Jev companion for Hermes Agent covering relevance, completion, recovery and optional admission decisions.
logicrw/awesome-jev-projects
Frequently asked questions
Who has final authority?
The main Hermes orchestrator and reviewer.
Can it use open models?
Yes, Laya and OpenJev backends are supported, including shadow comparisons.
Is it stable?
0.3.1 is the stable release; main contains newer development work.
Related projects
Local agent workbench whose decision crate borrows the System One state/questions request shape, starting with Cloudflare's Clef.
LibrarybetaRustSoftware-factory supervisor where Jev independently judges coding agents' completion, test sufficiency, progress and need for a human.
ProjectbetaPythonAsynchronous System-1 supervisory plugin for Hermes Agent with typed Jev decisions, ranking, verification and an opt-in tool gate.
ProjectbetaPythonHermes Agent skill that turns an intention into a checkable finish line, with shell checks and Jev judging the fuzzy requirements.
Agent skillprototypePythonHermes skill that calls Jev as a typed tiebreaker, returning Choice, Score or Noul probabilities so the agent can act or escalate.
Agent skillbetaPythonClaude Code Stop hook that blocks an unverified done: one four-question Jev call only when files changed and no check passed.
PluginbetaTypeScript