Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

goodvibes-jev

Rebuild of the goodvibes agent platform where Jev is targeted to make every semantic decision, with outcomes act, revise, defer or reject.

Projectprototype
View on GitHub ↗

Overview

goodvibes-jev rebuilds the goodvibes platform (SDK, daemon, TUI, agent and web UI) as one Bun monorepo. The stated target is for Jev to make every semantic decision, replacing keyword lists, regexes over prose, length cutoffs and guessed classifications, while preserving existing behaviour except where a newer autonomous decision contract supersedes it.

That contract, dated October 3, sets target requirements rather than claiming they are all implemented: Jev makes semantic decisions everywhere with no human in the runtime loop, with shared outcomes act, revise, defer and reject. A single judgment-port transport owns Jev unavailability, retrying or failing over and returning a typed unavailable result; products must not add retry loops, approval prompts or synthesised answers during outages. Code still enforces authentication, scope, revocation, fixed formats, monetary arithmetic and idempotency, and a Jev receipt cannot grant authority.

What the decision model decides

TypeQuestionAnswers
ChoiceFor this semantic decision point, should the system act, revise, defer or reject?act · revise · defer · reject

How it works

  1. A product reaches a semantic decision point
  2. The judgment port asks Jev
  3. Act on act/revise/defer/reject
  4. Code enforces authority, formats and arithmetic regardless of Jev

Strengths

  • Consistent decision contract across products
  • Clear separation of semantic judgment and hard enforcement
  • Explicit outage behaviour

Caveats

  • Requirements are targets, not all implemented
  • No human in the runtime loop by design

Getting started

See goodvibes-jev-intent.md and docs/design/autonomous-jev-decisions.md in the repository.

At a glance

Type
Project
Status
prototype
Language
TypeScript
Runtime
Bun monorepo (daemon, TUI, agent, web UI)
Has code
yes

Use cases and integrations

How curated lists describe it

Frequently asked questions

What does Jev decide?

Every semantic decision, with outcomes act, revise, defer or reject (as a target).

Can Jev grant permissions?

No, code still enforces authentication, scope and other hard boundaries.

What happens if Jev is down?

A shared transport retries or fails over and returns a typed unavailable result.

Related projects

  • oh-my-pi

    Oh My Pi coding agent (a Pi fork with IDE wiring) with an optional TypeSafe judgment provider for bounded decisions.

    ProjectproductionRust
  • AutoGPT

    AutoGPT agent platform, listed as adding Jev decision blocks for typed routing, filtering, scoring and confidence-gated next actions.

    ProjectproductionPython
  • oh-my-claudecode

    Multi-agent orchestration layer for Claude Code with opt-in Jev hooks for fast judgment points, decision caching and egress controls.

    PluginproductionTypeScript
  • SystemOneHarness★ 193

    Agent harness that compiles an environment's finite actions into typed questions for Jev and gates each step by confidence.

    CLIbetaPython
  • compact-adviser

    Agent plugin that asks Jev whether the session is at a safe boundary to /compact and can trigger it on Pi and Claude Code.

    PluginbetaTypeScript
  • dsh-jev-advisor

    DeepSeek Harness plugin that gives the human a Jev second opinion beside the agent's multiple-choice questions.

    ProjectbetaTypeScript