Jev by Example
Ten runnable JavaScript lessons where Jev judges agent situations like memory conflicts and handoffs while code enforces explicit policy.
Overview
Jev by Example from Really Artificial is a set of ten runnable experiments in building agents that know what to do next and when to ask. Each combines a small Jev judgment with readable application logic, focusing on decisions between agent steps: should a new memory replace an old one, did a successful tool call actually finish the task, is a retry about to create a duplicate invoice, is an allowed tool being used for an unrelated purpose, and are handoff constraints met.
The initial release has 10 examples with 34 authored cases. By default it runs in fixture mode with clearly labelled hand-authored responses, needing no install, account, key or network; live calls are opt-in. The HTTP implementation has been checked against the official API reference, but live model behaviour has not yet been verified. One example shows that a weekend-only Python exception should be kept as a scoped memory rather than replacing a standing TypeScript preference. It uses JavaScript on Node 20.17+ with zero dependencies and is MIT licensed.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | How should this new memory relate to the existing one (replace, keep scoped, ...)? | keep-scoped · propose-replacement · ... |
| Yes / no | Did the tool result actually complete the task?(inferred) | yes · no |
| Yes / no | Is this allowed tool being used for an unrelated purpose?(inferred) | yes · no |
How it works
- Pick an example
- Run in fixture mode with authored responses
- Inspect the question, inputs and policy
- Optionally switch to live Jev calls
Strengths
- Runs offline with no key
- Policies are explicit and editable
- Covers subtle agent edge cases
Caveats
- Live model behaviour not yet verified
- Fixture responses are hand-authored
Getting started
`git clone https://github.com/ReallyArtificial/jev-by-example.git && cd jev-by-example && npm start -- list`.At a glance
- Type
- Demo
- Status
- prototype
- Language
- JavaScript
- Runtime
- Node 20.17+
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Agent development: runnable JavaScript lessons use Jev Choice, Score, and Noul judgments for memory reconciliation, recovery proposals, and handoff checks, with explicit application policies, offline fixtures, and opt-in live calls.
yibie/awesome-jevTen runnable Jev examples for agent decisions: memory conflicts, tool-result checks, recovery, context selection, and handoffs. JavaScript, zero dependencies.
logicrw/awesome-jev-projectsTen runnable JavaScript lessons where Jev judges memory relationships, task evidence, and handoff constraints while application code enforces explicit policies; offline fixtures are the default and live calls are opt-in.
cobanov/awesome-jevTen runnable Jev examples for agent decisions: memory conflicts, tool-result checks, recovery, context selection, and handoffs. JavaScript, zero dependencies.
hellogumbo/awesome-jev
Frequently asked questions
Do I need a TypeSafe key?
No for fixture mode; live calls are opt-in.
What kinds of decisions are covered?
Memory reconciliation, tool-result checks, recovery, context selection, permission drift and handoffs.
Has it been tested against live Jev?
Not yet, per the README.
Related projects
Production control plane for OpenClaw agents: Jev routes each Slack message to its task, with Temporal workflows, an evaluator and memory.
LibraryproductionPythonX walkthrough of the Jev API followed by three demos: voice-controlled browsing, AI memory and YouTube preprocessing.
DiscussiondiscussionPortuguese agent skill teaching Claude Code, Cursor and Codex to use Jev correctly, bundled with the full raw research behind each claim.
Agent skillproduction- fast-jev-compaction★ 7,616
Claude Code plugin and npm library that replaces the compaction summary with Jev keep, truncate or drop scores for every tool call.
PluginbetaTypeScript MCP server giving coding agents fast typed decisions (ship gates, risk checks, file triage, page checks) as accept, review or escalate.
MCP serverbetaTypeScriptUnofficial Claude Code plugin that checks 'done' claims and small judgment calls with Jev, logs receipts, and can warn on unverified stops.
Pluginbeta