User repository on GitHub
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Demos & Experiments
Jev × Claude Code POC
Two Claude Code hooks powered by Jev, TypeSafe's System One model: a PreToolUse risk gate that asks three noul questions per tool call and maps the top probability to allow, ask or deny, plus a tier router that rewrites subagent calls to haiku, sonnet or opus.
- Stars
- 0
- Language
- Python
- Reading time
- 2 min
- Published by
- Community
- Author
- spmamidi
- Last verified
- Oct 2026
Highlights
- PreToolUse risk gate on Bash, Write, Edit, MultiEdit and NotebookEdit asks Jev three noul questions: destructive, outward-facing, credential-touching.
- Maps the top probability to allow, ask or deny and fails open on any error.
- Tier router asks one Choice over haiku, sonnet, opus for Task and Agent, rewriting the call via updatedInput.model; low confidence falls back to sonnet.
- Jev replaces bounded judgments — router and judge calls, JSON-schema retry loops — never the agent's reasoning.
- README rule of thumb: route whole work units (sessions/subagents), not mid-conversation prompts, because mid-loop steering busts the prompt cache.
Watch out
README and GitHub metadata reviewed 2026-10-17; the digest truncates the quick start after the pip install line and no licence is listed, so setup beyond installing typesafe-sdk and any measured results are not captured.
Questions about this entry
Short answers drawn from the linked project or page.
Two Claude Code hooks powered by Jev: a PreToolUse risk gate for tool calls, and a tier router that chooses the model for subagent briefs.
It asks Jev three noul questions per tool call — is it destructive, outward-facing, credential-touching — and maps the top probability to allow, ask or deny, failing open on errors.
It asks Jev one Choice question over haiku, sonnet and opus for the subagent brief and rewrites the call via updatedInput.model; a low-confidence pick falls back to sonnet rather than downgrading.
pip install typesafe-sdk for live mode, or omit it for mock mode, per the quick start; the digest cuts off right after that line, before keys or configuration are explained.
The agent's reasoning. The README says Jev replaces bounded judgments such as router and judge calls and JSON-schema retry loops.
Route whole work units — sessions or subagents — rather than mid-conversation prompts, because mid-loop steering busts the prompt cache, per the README's rule of thumb.
The README describes calibrated Choice, Score and Noul decisions in one forward pass, in about 70–500 ms, at a fraction of a frontier-LLM call's cost.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
YouTube: Ray Amjad puts Jev inside an agentic coding loop and looks at what it costs to run.
YouTube: RoboNuggets puts Jev to work inside Claude Code.
Reddit: A Claude Code mod replaces the auto-mode classifier with Jev, 2x faster.
More Jev resources like this
An MIT-licensed case study with two working Python programs pairing Jev and Claude Code: a support-triage app with Jev inside it, and a Claude Code safety hook where Jev judges the agent's own shell commands. Tests and evals run offline, free.
Jev Practices & Patterns#community#github#python
A free Claude Code skill that puts Jev in front of the expensive model: thirteen commands sort, check, score, gate and verify with Jev, so Claude reads only the unsure slice and writes the output. Needs python3 and an OpenRouter key.
Jev Agents & MCP#community#python#claude-code
Community
charlesdove977/claude-x-jevPython
charlesdove977/claude-x-jevPythonA small SeleniumBase project that sends a page's reduced HTML to Jev and gets back three structured answers: whether a CAPTCHA is present, a 0-100 confidence score and which of six types it is. The README warns scores vary between runs and are not calibrated probabilities.
Jev Demos & Experiments#community#github#captcha
Community
mdmintz/jev-testingPython
mdmintz/jev-testingPythonDemo pack from the French-language Jev meetup of 8 October 2026 on System One models: a Chrome ad-blocker, an unclutter browser fork, a decision-routing kit comparing a local engine with Jev via OpenRouter, and n8n email triage.
Jev Demos & Experiments#community#github#demos
Community
royam0820/Jev-MeetupJupyter Notebook
royam0820/Jev-MeetupJupyter NotebookA runnable reference implementation of semantic data quality: it catches silent meaning drift that passes every deterministic BigQuery check, treats a semantic quality score as a first-class ops metric, and puts a circuit breaker at the gate.
Jev Demos & Experiments#community#github#python
A Python dispatcher skeleton for TypeSafe System One (Jev) that turns typed signatures into batched questions, plus a Playwright browser harness that measures the dispatcher against ground truth.
Jev Demos & Experiments#community#github#python
Community
monkey1sai/jev-systemone-labPython
monkey1sai/jev-systemone-labPythonFrom 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.