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

jev-map

Maps each Python function to the tests that run it; Jev scores candidate function-test links that static analysis misses.

ProjectbetaMIT
View on GitHub ↗Live demo ↗

Overview

jev-map builds a test map for Python repositories, aimed at coding agents that need to know which tests exercise a given function. Structural links come from reading the code without running anything, following resolvable calls, import aliases and similar patterns. Optionally, Jev adds inferred links that code reading misses.

For a function, keyword search picks a few likely tests and Jev scores each function-test pair from 0 to 1. Pairs at 0.8 or higher are added as inferred links with a receipt; lower pairs are left out. Every link records whether it came from the code or from Jev, so an agent can decide how much to trust it. The list summary says inferred links are checked against real test runs. The tool runs locally by default, prints JSON from every command and also ships an MCP server.

What the decision model decides

TypeQuestionAnswers
Yes / noDoes this test exercise this function?(inferred)yes · no

How it works

  1. Run refresh to build structural links from the code without executing it
  2. For each function, keyword search proposes about three candidate tests
  3. Jev scores each function-test pair from 0 to 1
  4. Pairs at or above 0.8 become inferred links with a receipt
  5. Agents query related-tests or explain-link via CLI or MCP

Strengths

  • Every link is labelled with its provenance
  • Fixed 0.8 cutoff keeps inferred links conservative
  • Core has no dependencies and runs locally by default
  • MCP server makes the map available to agent hosts

Caveats

  • Inferred links need a TypeSafe key and send source excerpts to the API
  • Map file holds source excerpts, so it should be gitignored
  • Python repositories only

Getting started

pip install "jev-map @ git+https://github.com/krishhgg/jev-map" then jev-map --repo /path/to/repo refresh

At a glance

Type
Project
Status
beta
Language
Python
Runtime
CLI and MCP server
Has code
yes
Has tests
yes
License
MIT

Use cases and integrations

How curated lists describe it

  • Map each Python function to the tests that run it. Jev adds the links code analysis misses, checked against real test runs.

    hellogumbo/awesome-jev

Frequently asked questions

What does Jev decide in jev-map?

It scores how likely a candidate test is to exercise a function; only pairs scoring 0.8 or higher are added.

Do I need Jev to use it?

No. Structural links work without any model; Jev is an optional layer for links static reading misses.

Does it run my code?

Building the map does not execute your code.

Related projects

  • jev-bugfix

    Codex skill using the local Jev CLI to rank candidate code snippets before reproduction, root-cause analysis and a verified fix.

    Agent skillprototypePython
  • agent-fastpath

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

    MCP serverbetaTypeScript
  • CodexQA Jev Browser

    CodexQA browser skill that indexes in-page controls and has Jev pick the operation and target each step, with no screenshots needed.

    Agent skillbeta
  • jev-kit

    Codex skill and MCP for advisory Jev decisions in Java/Maven backend work: file relevance, change risk, SQL risk and more.

    MCP serverbetaTypeScript
  • jev-mobile-mcp

    Fork of mobile-mcp that adds OCR and lets Jev pick which recognised text element to tap on Android and iOS devices.

    MCP serverbetaTypeScript
  • jev-sim-wingman

    Single-file MCP server giving coding agents typed, budgeted Jev control over the iOS Simulator with gated destructive actions.

    MCP serverprototypeJavaScript