Jevidence
Python sandbox turning Jev issue-category and reproduction judgments into testable routing proposals, with offline replay.
Overview
Jevidence is a companion to a Peak Evergreen article on Jev for developers. It is a Python sandbox for turning Jev judgments into testable issue-routing proposals: an issue goes to Jev for typed judgments, your policy turns them into a proposed queue with review hints, and on failure the existing queue is kept. In opt-in live mode it asks Choice and Noul questions about issue category and whether reproduction steps are present, then applies confidence and reproduction gates; it proposes rather than assigns.
Offline commands need no packages, key or network: a demo shows the same synthetic judgment falling back at a 0.8 route floor and proposing runtime investigation at 0.7, an evaluate command runs policy checks, and replay reads stored judgments from JSONL with independent expected-queue labels to report wrong specialist routes, review volume, unavailable judgments, coverage and disagreements at several thresholds without ever contacting a model. The author notes it was substantially built with Codex and is independent of TypeSafe. It uses only Python 3.10+.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which category does this issue belong to? | — |
| Yes / no | Does this issue include reproduction steps? | — |
How it works
- Ask Jev about category and reproduction (live mode, opt-in)
- Apply confidence and reproduction gates
- Propose a queue or fall back to review
- Replay stored judgments against labels at different thresholds
Strengths
- Offline demo and replay with no network
- Separates judgment from policy
- Measures threshold trade-offs on labelled data
Caveats
- Sandbox for learning; does not assign issues
- Synthetic fixtures by default
Getting started
git clone, then `python3 -m jevidence demo` and `python3 -m jevidence evaluate`.At a glance
- Type
- Project
- Status
- prototype
- Language
- Python
- Runtime
- CLI
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
Developer education: Python sandbox asks Jev `Choice` and `Noul` questions about issue category and reproduction steps in opt-in live mode, then applies confidence and reproduction gates to propose a queue or review fallback without assigning the issue, with synthetic offline fixtures and policy tests.
yibie/awesome-jev
Frequently asked questions
Does it assign issues?
No, it proposes a queue and review hints; failures keep the existing queue.
Can I test thresholds without calling Jev?
Yes, replay uses stored judgments and never contacts a model.
What do I need to run the demo?
Only Python 3.10+; no packages, key or network.
Related projects
Runnable Jev cookbooks, a plain-English guide and sixteen interactive simulations of how a typed decision model changes architecture.
AppbetaPythonGitHub Action that triages issues with Jev, applying label, spam, needs-info and duplicate tags only above set thresholds.
ProjectbetaTypeScriptOne YAML task to classify, route, score or verify text with rules, Jev, NanoJev or an OpenAI-compatible model, with cost records.
CLIbetaPythonMultiplayer GitHub triage dashboard where Jev answers kind, severity, urgency, duplicate and next step per issue.
AppbetaCLI, MCP server and skill for issue triage where Jev predicts issue kind, affected module and whether information is sufficient.
MCP serverprototypePythonTwo beginner Python apps using Jev: a message sorter and a support ticket router that send uncertain items to human review.
ProjectprototypePython