User repository on GitHub
WorkBuddy UserPromptSubmit hook: judge each prompt with TypeSafe System One before the model sees it, to cut main-model token spend.
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Integrations & Adapters
Jev Pre-flight Gate
UserPromptSubmit hook for WorkBuddy (and a pre_llm_call shell hook for Hermes Agent) that has Jev judge every prompt in about a second and inject the verdict as routing hints, aiming to cut main-model tool calls and turnarounds.
- Stars
- 1
- Language
- Python
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- adamwdff
- Last verified
- Oct 2026
- Languages
- Python
- Capabilities
- Open sourceRequires API key
Highlights
- Judges each prompt before the main model reads it and injects cues like 'do not ask to confirm', aiming to stop wasteful tool calls.
- Injects a verdict line with dimensions such as task_type, needs_files, ambiguity, risk, scope, depth and skill_route, plus routing hints.
- Fail-open by design: no key, timeout, error or empty answer passes the prompt through; three consecutive failures trigger a 10-minute circuit breaker.
- The README measures about 300-400 tokens injected per turn, a net loss near 1% on simple requests and positive in expectation overall.
- Key lives at ~/.workbuddy/.typesafe_key with 0600 permissions; /jev status, /jev on and /jev off are self-consumed and cost zero model tokens.
Quickstart
bash
git clone https://github.com/adamwdff/jev-gate.git && cd jev-gate && python3 install.pyWatch out
Copy reviewed against the repo README on 2026-10-15; the README is mostly Chinese and the excerpt covers the WorkBuddy path plus the Hermes differences. The 300-400 token cost and 1% loss estimates are the author's own measurements, not an independent audit.
Questions about this entry
Short answers drawn from the linked project or page.
A WorkBuddy UserPromptSubmit hook and a Hermes Agent shell hook that let Jev make a structured judgment on each request in about a second before the main model sees it, then inject the verdict as routing hints.
Main-model spend driven by unnecessary tool calls (a single grep can cost 5k-50k tokens), extra round trips and uncontrolled output length. Blocking one avoidable grep earns back thousands of tokens, per the README.
A line such as [Jev pre-flight] with task, history, files, correction, ambiguity, risk, scope, depth and route values, plus cues like 'do not ask to confirm, act now'. The README also shows a Jev echo for the user.
The README measures roughly 300-400 tokens of injection cost per turn, a net loss of about 1% on requests that would never read files; positive in expectation, but not profitable on every message, per the author.
Fail-open is the rule: no key, timeout, error or empty response exits silently and lets the prompt through, never blocking questions. After three consecutive failures a circuit breaker pauses API calls for 10 minutes.
In ~/.workbuddy/.typesafe_key for WorkBuddy or $HERMES_HOME/.typesafe_key for Hermes, with 0600 permissions, separate from settings.json and config.yaml and gitignored. The README warns logs include the first 120 characters of prompts, so do not share them.
Clone the repo and run python3 install.py for WorkBuddy or python3 hermes/install_hermes.py for Hermes, then fully restart the app or start a new session. /jev off disables it without uninstalling, and /jev on re-enables it.
More Jev resources like this
Two opt-in Hermes Agent plugins that add TypeSafe Jev decision layers: a pre-tool gate for terminal, write_file, and patch calls, and a skill router that names the one skill worth loading before the model call. Stdlib only and fail-open.
Jev Integrations & Adapters#community#catalog-import#logicrw
Community
dogmatiic/hermes-jevPython
dogmatiic/hermes-jevPythonEffortlane is a local router for Codex on macOS that picks model and reasoning effort with Jev before each user turn, keeps native ChatGPT authentication, and records routing evidence locally. Shadow mode logs recommendations without changing execution.
Jev Integrations & Adapters#community#catalog-import#logicrw
A Hermes Agent plugin that adds per-turn model routing — Jev picks cheap, primary or complex before each call — and a trim-then-compress context engine that forbids compressing without trimming first.
Jev Integrations & Adapters#community#catalog-import#logicrw
Community
punkcanyang/hermes-jev-routerPython
punkcanyang/hermes-jev-routerPythonAn opencode plugin that turns an agent's small decisions — is this command safe, which tool fits, is this PR ready — into typed Jev questions with calibrated confidences, labelling each answer auto or escalate so the agent knows when to ask for help.
Jev Integrations & Adapters#community#catalog-import#logicrw
Community
darwintechlab/openjevJavaScript
darwintechlab/openjevJavaScriptA GitHub App that routes pull-request review work through Jev: it ingests the webhook, normalizes the PR context, gets a LOW, NORMAL or SPECIALIST decision, and a deterministic policy layer maps it to auto-approval, a review request or a check-run update.
Jev Integrations & Adapters#community#catalog-import#logicrw
Community
hexyedev/jevprPython
hexyedev/jevprPythonAn OpenCode plugin that keeps the session on a cheap sticky parent model so the cache stays warm, and asks Jev whether a turn is hard before escalating through jev_escalate to temporary stronger subagents.
Jev Integrations & Adapters#community#catalog-import#cobanov
From 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.