User repository on GitHub
Connect TypeSafe Jev to Grok Bot as a cheap decision layer - usage gates, skill template, examples
- Forks
- 13
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
Jev directory
CommunityJev Tools & CLIs
Grok Bot Jev Router
A reference router and skill that put Jev in front of Grok Bot so each request is classified before expensive research, browser, retry, or subagent work, returning actions such as reuse_cache or ask_human.
- Stars
- 112
- Language
- Python
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- Bodila51
- Last verified
- Sep 2026
About this project
Connect TypeSafe Jev to Grok Bot as a cheap decision layer. Jev classifies the request before expensive research, browser, retry, or subagent work, so Grok Bot can reuse a fresh artifact, stop a failing retry, cap research, or ask for approval. This package is a small open-source reference implementation. It does not change Grok Bot's foundation model and does not route Cursor models. Visualization of Grok Bot + Jev at work: A Python router around the TypeSafe SDK's systemone call.
Quoted from the project's README. Read the full README.
Works forJev for Routing
Used inJev for Software & DevTools
Highlights
- Explicit actions include reuse_cache, stop_retry, run_deterministic, chat_only, research_capped, allow_subagent, and ask_human, with shadow and active modes.
- A kill switch disables the router through enabled: false or a bypass jev marker, and src/secrets.py reads only TYPESAFE_API_KEY from the environment.
- In a recorded local A/B across five tasks, flight browser opens fell from 1 to 0, same-approach retries from 3 to 0, and research pages from 10 to 4.
- A separate 24-candidate timing run took 53.803 s for the uncapped arm and 4.125 s for the Jev-ranked top-five arm, with two Jev calls at about $0.000405.
- It never sends, publishes, pays, deletes, or changes permissions without a human approval path, and it logs decisions without the API key.
Quickstart
bash
python3 -m venv .venv
.venv/bin/pip install -r requirements.txt
cp config.example.yaml config.yaml
export TYPESAFE_API_KEY=YOUR_API_KEY
.venv/bin/python scripts/dry_run.pyWatch out
MIT-licensed. Needs Python and a TYPESAFE_API_KEY; shadow mode enforces nothing, active mode depends on the pasted skill honoring decisions, and the A/B figures are one local run's proxy measurements.
Questions about this entry
Short answers drawn from the linked project or page.
No, and the repo does not claim it. The included data is one local run's proxy measurements: the Grok Bot weekly usage meter moved 37% to 38% and then 39% after the second phase, and exact per-task Grok tokens were unavailable, so it is explicitly not a token-savings claim. The count-based numbers (browser opens 1 to 0, retries 3 to 0, research pages 10 to 4) are the honest part.
The Jev arm capped research to the ranked top five of 24 candidates, so fewer page fetches were by design rather than an efficiency result. That run took 4.125 s and 5 pages against 53.803 s and 14 pages for the uncapped arm, using two Jev calls at about $0.000405.
Only if the pasted skill honors route.action. Shadow mode logs decisions and enforces nothing, and this is not a hidden interceptor, so a bot that ignores the skill keeps running. There are also no pre-wake hooks here, so the cost of waking Grok Bot in the first place is untouched.
No. It returns an action such as reuse_cache, stop_retry, research_capped, or ask_human; it does not execute browser actions, research, payments, messages, or account changes. By design it never sends, publishes, pays, deletes, or changes permissions without a human approval path, and it logs decisions without the API key.
Clone the repo, make a venv, pip install -r requirements.txt, copy config.example.yaml to config.yaml, and export TYPESAFE_API_KEY. scripts/dry_run.py sends the sample states through Jev and prints JSON decisions. For a no-network check, set enabled: false in config.yaml first, which disables the router while the same script still runs.
src/secrets.py reads only TYPESAFE_API_KEY from the environment and never a secrets file, and the README asks you to keep config.yaml, .env and local logs uncommitted. The router is a Python wrapper around the TypeSafe SDK system_one call; the Grok Bot skill passes only task metadata and short context, never credentials.
Separate, and deliberately small. It is a reference implementation made of a Python router plus a skill file, skill/jev-usage-router.SKILL.md, that you copy into a Grok Bot skill. It does not change Grok Bot's foundation model and does not route Cursor models. The whole repository is MIT licensed.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Facebook: Intake Router keeps diagnosis out of the answer space at the front desk.
FacebookClinicnotes
X: OpenRouter Decision Model Rankings
X: herdr + jev, an agent router
X: A group chat bot with decision room routing
More Jev resources like this
A Python toolkit that installs nine plain SKILL.md files for Hermes, Claude Code, and Codex, letting Jev route models, filter retrieved passages, select skills, and choose bounded computer or browser actions.
Jev Tools & CLIs#community#python#agents
Community
kerpopule/hermes-jev-skillsPython
kerpopule/hermes-jev-skillsPythonLangChain's langchain-typesafe integration exposes Jev as a TypeSafeClassifier plus experimental model-routing and auto-mode middleware that screens risky tool calls before execution.
Jev Tools & CLIsArticle#community#integration#python
Community
A Go-based team workspace with optional Jev AI for agent auto-routing, task auto-fill, auto-assign, and automation condition nodes.
Jev Tools & CLIs#community#go#agents
Community
Paca-AI/pacaGo
Paca-AI/pacaGoA collection of Jev demos, workflows, and agent skills for coding agents: five installable skills and 108 scenarios, installed by pointing Codex, Claude Code, or OpenCode at an agent prompt.
Jev Tools & CLIs#community#python#skills
Community
wuyoscar/jev-skillPython
wuyoscar/jev-skillPythonA TypeScript layer beside an LLM: JevSift selects approaches, JevLoop audits progress with typed actions, and JevLong watches sessions for stalls without steering them.
Jev Tools & CLIs#community#typescript#coding-agent
Community
Alex314618-create/JevRevTypeScript
Alex314618-create/JevRevTypeScriptA Python harness for System One models. It runs Jev and other decision models locally, or through HarnessRouter.ai.
Jev Tools & CLIs#community#python#open-models
Community
HarnessRouter/SystemOneHarnessPython
HarnessRouter/SystemOneHarnessPythonFrom 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.