Skip to content

jevpipe

Unix filter CLI that pipes files, lines or records to Jev for bulk yes/no, choice or score judgments so agents need not read everything.

Projectbeta

Overview

jevpipe gives a coding agent a 'System 1': fast, cheap judgments over thousands of files, lines or records in one shell command. Each item is judged on its own. The filter subcommand keeps items for which a yes/no question is true, and map returns typed answers such as a choice or score per item. A summary on stderr reports records kept, skipped and failed, tokens, cost and time.

The README's example filters 19 ripgrep commit messages for new features in about a second for a fraction of a cent, and another applies questions to 1,000 banking support messages. Suggested uses include searching code by meaning, triaging CI failures and similar bulk tasks. It works with an OpenRouter key by default or a TypeSafe key, stored in the system keychain or set by environment variable. An agent skill teaches coding agents when and how to call it.

What the decision model decides

TypeQuestionAnswers
Yes / noIs this a new feature? (any per-item yes/no via filter)—
ChoiceIs this CI failure a timeout, network error, failed assertion or crash?timeout · network error · failed assertion · crash
Yes / noDoes this file retry failed requests?—

How it works

  1. Pipe items (lines, files or records) into jevpipe
  2. Each item is sent with the question to Jev
  3. filter keeps items above the yes threshold; map emits typed answers
  4. Summary of counts, tokens and cost goes to stderr

Strengths

  • Bulk judgments in one shell command
  • Agent skill for coding agents
  • Cost and token summary per run
  • Works with OpenRouter or TypeSafe

Caveats

  • Each item must be answerable on its own
  • Every item is a paid call
  • Requires an API key

Getting started

uv tool install jevpipe && jevpipe auth set-key && npx skills add fabianboth/jevpipe --skill jevpipe

At a glance

Type
Project
Status
beta
Language
Python
Runtime
CLI and agent skill
Has code
yes
Has tests
yes

Use cases and integrations

How curated lists describe it

  • Pipe anything into Jev, get typed decisions out. A Unix filter that lets agents offload bulk judgments to a System One model.

    hellogumbo/awesome-jev

Frequently asked questions

What is the difference between filter and map?

filter keeps items where a yes/no question is true; map returns a typed answer per item.

Which keys does it accept?

An OpenRouter key by default, or a TypeSafe key after setting the provider.

How do agents learn to use it?

Install the provided skill with npx skills add.

Related projects

  • Claude Code plugin that finds fuzzy decision points in an app and ranks where a Jev question could replace them, with precedents.

    ProjectbetaPython
  • Skill and MCP service that audits an app for AI calls that are really decisions and shadow-tests migrating them to Jev.

    MCP serverbetaTypeScript
  • Claude Code plugin and Python CLI that hands judgments repeated over many items to Jev, with redaction, rate limits and a pilot run.

    Agent skillbetaPython
  • Claude Code and Codex skill that routes a task to one of 27 specialist roles, with an optional Jev engine for catalog selection.

    ProjectbetaPython
  • MCP server giving coding agents fast typed decisions (ship gates, risk checks, file triage, page checks) as accept, review or escalate.

    MCP serverbetaTypeScript
  • Deterministic code-graph context for coding agents on Postgres, AGE and pgvector, with optional Jev or local re-selection of symbols.

    MCP serverbetaPython