User repository on GitHub
Codex-first CLI for bounded Jev judgments without echoing input state
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Agents & MCP
Jev Judge — ohmyjiro
Dependency-free Python 3 CLI that hands a JSON state plus 1 to 16 named Noul, Choice or Score questions to Jev and prints typed answers, usage and elapsed time without echoing the state back.
- Stars
- 0
- Language
- Python
- License
- MIT
- Published by
- Community
- Author
- ohmyjiro
- Last verified
- Oct 2026
This is one of several Jev listings for the same project. The canonical entry is Jev Judge — 00200200.
About this project
A small, dependency-free CLI for delegating bounded judgments to TypeSafe Jev. Give it a JSON state and named Noul, Choice, or Score questions. It returns typed answers, usage, and elapsed time without echoing the state.
Quoted from the project's README. Read the full README.
- Languages
- Python
Highlights
- The state never round-trips: output carries answers, model, usage and elapsedMs, so the agent reads a small decision instead of the full evidence.
- Bounded by design, at 1 to 16 named questions per request and a 50 KB request guardrail, with related questions about one state kept in a single call.
- The documented agent pattern builds the request from a local file or tool output in code, before that material enters the agent's context.
- Nothing to install: clone, install -m 755 bin/jev-judge into your bin directory, and supply TYPESAFE_API_KEY from a secret manager.
Watch out
The context saving is a real pattern, but the README is explicit that it does not make the TypeSafe API call free or prove a workflow is faster. The CLI chooses no confidence threshold and takes no action; both are yours. There is no batch mode: for many independent records you call it once per record from code, so the request count scales with your record count. The CLI sends the supplied state and questions to TypeSafe's API, and the README says not to submit confidential material unless that transfer is appropriate for your task. The key must come from the environment or your secret manager and never from a request file or the repository. Jev does not write summaries or code, so open-ended writing and final verification stay with the agent. It is an independent community tool, not affiliated with TypeSafe AI. No cost per call, latency figure, accuracy number, or rate limit is published; the bundled example is one saved link with three questions, and the project's own check is a unittest run. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
The mechanism is real but the benefit is unquantified. You build the request from a local file or tool output in code, before that material enters the agent's context, and the agent then reads the small answer instead of the full evidence. What the README refuses to claim is that this makes the TypeSafe call free or that the workflow is faster. No token-savings figure, latency number or cost per call is published, so you would be estimating the gain before committing to the pattern.
One call per record for many independent records, which the README states as the pattern: call once per record from code. There is no documented batch or concurrency option, so 500 records means 500 requests and your own rate-limit handling. What it does optimize is the other direction: related questions about the same state belong in one request, bounded at 1 to 16 named questions and a 50 KB request guardrail.
The key comes from TYPESAFE_API_KEY through your existing secret manager, and the README is explicit that it must not go in a request file or be committed. On the state: it does leave your machine, because the CLI sends the supplied state and questions to TypeSafe's API, so the README says not to submit confidential material unless that transfer is appropriate. On output, the answer side never includes the input state, and API failures are reported without printing the state or the key.
You get a reported API failure, not a crash, and neither the state nor the key is printed. That is as specific as the README gets: it documents no error codes, no retry behaviour, no backoff and no exit-code contract, so you would be matching on message text rather than on typed failures if you script around it. The tool also does not choose a confidence threshold for you, so uncertain answers are yours to route.
The output contains answers, model, usage and elapsedMs, and never includes the input state, so an agent reads the small decision rather than the full evidence. Pipe it through jq '.answers' if you only want the decisions.
One call accepts 1 to 16 named questions and enforces a 50 KB request guardrail. Related questions about the same state belong in one request, while many independent records need one call each.
No. It reports answers without choosing a confidence threshold or taking an action, and Jev does not write summaries or code, so open-ended writing and final verification stay with the agent.
More Jev resources like this
51GitHub stars
Local LinkedIn recruiting workspace on Browser Use's Jev Ultrafast: Jev screens titles, drives navigation, and assesses profile excerpts against your brief while a live browser stays visible; matches and evidence export as JSON.
Jev Agents & MCP#community#jevbest#catalog-import
Community
skeptrunedev/jev-recruiterPython
skeptrunedev/jev-recruiterPython33GitHub stars
Codex Skill plus stdlib-only Python CLI that recommends which installed skill to read next for a task, allowing no-skill or uncertain answers and never executing candidates.
Jev Agents & MCP#community#jevbest#catalog-import
Community
win4r/jev-skill-suggesterPython
win4r/jev-skill-suggesterPython72GitHub stars
MCP server over stdio that gives coding agents fourteen Jev judgment tools, turning evidence plus an enumerable question into auto, review, or escalate actions from calibrated probabilities.
Jev Agents & MCP#community#jevbest#catalog-import
Community
PyModel/jev-judge-mcpPython
PyModel/jev-judge-mcpPython21GitHub stars
Research runtime that pages a large logical tool and context space through a small bounded Jev decision surface with explicit approvals.
Jev Agents & MCP#community#jevbest#catalog-import
25GitHub stars
Agent-oriented playbook for using Jev as a cheap, fast decision layer that triages, classifies, scores, and gates work before expensive browser, research, retry, or subagent steps.
Jev Agents & MCP#community#awesomejev#catalog-import
Community
Bodila51/muse-jev-playbookPython
Bodila51/muse-jev-playbookPython14GitHub stars
Opt-in Hermes plugin that routes each turn to one skill: a wide Jev Choice plus request checks gate a narrow fits round, injecting a single ignorable hint line.
Jev Agents & MCP#community#jevbest#catalog-import
From the Jev community
Posts from builders shipping with Jev right now.