What it is
A small open-source Python reference implementation that connects TypeSafe Jev to Grok Bot as a cheap decision layer, classifying requests before expensive research, browser, retry, or subagent work. For developers integrating Jev with Grok Bot.
How it uses Jev
Jev classifies the request state before expensive work. The router calls the TypeSafe SDK's system_one and returns actions like reuse_cache, stop_retry, run_deterministic, chat_only, research_capped, allow_subagent, or ask_human. In shadow mode decisions are logged and advisory; in active mode the Grok Bot skill honors route.action.
Primitives:choice
Technique worth stealing
Use a cheap classification call to gate expensive tool work, with shadow mode for measurement before active enforcement.
Try it
git clone the repo, install requirements, set TYPESAFE_API_KEY, then run scripts/dry_run.py.
Evidence
Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.
- Jev-centricyes0.84
- Shows a System One patternyes0.64
- Handles uncertaintyno0.10
- Measuredno0.06
- Runnableyes0.87
- Worth recommendingyes0.61
- Model replicano0.03
- Problem scopescore on a 0–2 scale1.06
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.