npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

kibi-plugin-jev

v0.3.8

Published

Optional TypeSafe Jev semantic classifier for Kibi

Readme

kibi-plugin-jev

Optional Kibi capability plugin that uses TypeSafe Jev (@typesafe-ai/sdk) for two capabilities:

  • kibi.semantic-classifier.v1 — lane classification for semantic advisor propositions.
  • kibi.vocabulary-alignment.v1 — modeling-time vocabulary convergence: choosing an existing subject for a new clause (rankSubjects) and flagging claims that may restate an existing one (compareClaims).

Each capability is activated separately. Activating one never activates the other.

Install and activate

npm install --save-dev kibi-plugin-jev
mkdir -p ~/.config/kibi
printf '%s\n' 'TYPESAFE_API_KEY=...' >> ~/.config/kibi/env

Optional project override: put TYPESAFE_API_KEY=... in <workspace>/.env.kibi. Process environment variables always win. No Cursor/OpenCode/Codex/ZCode-specific secret config is required — any harness that starts kibi / kibi-mcp uses the same resolution. Restart long-running MCP after changing env files.

Activation is explicit in the project package.json. Installing the package does not activate it, and the API key must not be stored in that file.

{
  "kibi": {
    "plugins": [
      {
        "package": "kibi-plugin-jev",
        "capabilities": {
          "kibi.semantic-classifier.v1": { "mode": "augment" },
          "kibi.vocabulary-alignment.v1": { "mode": "augment" }
        }
      }
    ]
  }
}

Restart long-running Kibi MCP or client processes after changing kibi.plugins. Removing that entry disables the plugin.

| Variable | Required | Meaning | | --- | --- | --- | | TYPESAFE_API_KEY | yes, when a call is made | TypeSafe credential. Prefer ~/.config/kibi/env or .env.kibi. Read only when a client is constructed for a real call. | | KIBI_JEV_MODEL | no | Model id. Empty is unset. Default jev-latest. | | KIBI_JEV_TIMEOUT_MS | no | Positive integer milliseconds, at most 120000. Malformed values throw JevProviderError (malformed) before any network call. |

augment lets builtin handle normal cases and asks Jev only for what builtin left unresolved (an unresolved or ambiguous lane; a clause the builtin ranker left as new_subject; a claim pair the builtin did not judge a duplicate). replace gives Jev the capability and uses builtin only after provider failure. shadow runs Jev for comparison and cannot change canonical output.

Vocabulary alignment

Kibi's builtin ranker always runs first and is complete on its own: it scores existing subject facts by IDF-weighted token overlap with the clause, the subject title, and the titles of requirements that already constrain the subject. Jev refines that result:

  • rankSubjects asks one choice question per clause over the builtin top 5 candidates plus new_subject. Jev can only pick among those candidates; any other answer is rejected and Kibi falls back to builtin.
  • compareClaims asks one noul question per claim pair ("do these two claims state the same obligation?"). A yes is a review candidate (review:possible-duplicate), never a verdict.

Vocabulary alignment runs only inside requirement modeling (the kb_model_requirement operation: MCP kb_model with mode: "requirement", CLI kibi model-requirement). kb_check and kibi check never call it: the Prolog checks (domain-redundancy, subject-key-identity, and the others) stay deterministic, offline, and the only pass/fail authority.

Behavior and safety

Explicit JevSemanticClassifierOptions (model, timeoutMs, apiKey, clientFactory) override the environment for both capabilities. That constructor (createJevPlugin, createJevSemanticClassifier, createJevVocabularyAlignment) is for programmatic embedding and tests. package.json does not accept provider options.

Installing the package without activation makes zero TypeSafe calls. Importing the module does not create a client. Kibi does not depend on Jev. When active, proposition or clause text (and, for subject ranking, candidate subject keys and requirement titles) is sent to TypeSafe. On provider failure Kibi falls back to the builtin provider and stamps the result fallbackUsed: true. Diagnostics include the effective model and never the API key. The host loads the named kibiPlugin export. kibi doctor reports each activated capability and mode, secret source labels, and Jev model/timeout without values.