Tribar Connector for TypeSafe Jev

Description

Most AI plugins hand you a paragraph of prose and leave you to parse it. This one does the opposite. It connects WordPress to the TypeSafe System One API, whose Jev model answers typed questions and returns values your code can branch on directly:

• Noul — the probability, from 0 to 1, that a statement is true.
• Choice — one option from a set you define, with the full probability distribution and a confidence figure.
• Score — a position on an ordered scale you define, with probabilities per level and a confidence figure.

Many questions can be sent in a single call and are evaluated in parallel.

In five lines

use JevConnector\Question;

$answer = JevConnector\ask(
    array( 'comment' => $comment->comment_content ),
    array( 'hostile' => Question::noul( 'Does this comment attack a person rather than an idea?' ) )
);

if ( ! is_wp_error( $answer ) && $answer->noul( 'hostile' ) >= 0.9 ) {
    wp_set_comment_status( $comment, 'hold' );
}

A probability you branch on, not prose you parse. For a choice or a score the
gate is is_confident(); for a noul the probability is itself the confidence,
so the threshold goes straight on the value.

At its core this is a connector, not a feature. The plumbing comes first:

• Registration with the WordPress Connectors API, so your key is managed by core under Settings → Connectors and never by a field this plugin invented.
• A PHP client with retry and backoff on rate-limit and overload responses.
• Question builders that validate the shape before anything leaves your site.
• A response reader with one accessor per answer type, plus a confidence gate.
• Response caching keyed by request content, so re-evaluating identical material costs nothing.
• A REST proxy at /wp-json/jev/v1/ask so admin-side JavaScript never sees your key.
• Filters and actions at every point worth intercepting.

This plugin adds no menu and no screens of its own. Like the provider connectors for Anthropic and Google, it is the card under Settings → Connectors plus a PHP and REST surface for other code to build on. What a site does with Jev’s answers belongs to the theme or plugin that owns the content.

TypeSafe and Jev are trademarks of their owner. This plugin is an independent client and is not affiliated with, endorsed by, or sponsored by TypeSafe, and it ships no TypeSafe logo.

Common uses

Spam is not one of them: leave that to Akismet, which sees signals from millions of sites. Jev answers questions about meaning that a spam filter does not ask.

• Comment moderation: “Does this comment attack a person rather than an idea?” Hold it for a human.
• Contact form routing: “Which team should answer this: billing, technical, sales, other?” Send it to that inbox.
• Tags and categories: “Is this post about this term?”, one question per existing term in a single call. Suggest the terms; an editor accepts.
• Pre-publish checks: “How well is the central claim supported?” on a scale you define. Show it in the publish checklist.
• Policy and compliance: “Does this post make a health claim?” Require review before publishing.
• Checking AI-written drafts: “Does the summary state anything the source does not?” Flag it before it goes out.
• Product listings: “Is this a prohibited item under our policy?” and “Which category fits?” Queue it for review.
• Reviews and feedback: “What is the tone?” and “Does this report a defect?” Route defects to support.
• Related posts and search: “How relevant is this post to the one being read?”, per candidate. Reorder the list.
• Lead qualification: “How well does this enquiry fit the customer we serve?” Answer the best fits first.

Anything that publishes, deletes or emails should wait for a high confidence or a human.

Example

$answer = JevConnector\ask(
    array( 'comment' => $comment_text ),
    array(
        'off_topic' => JevConnector\Question::noul( 'Is this comment unrelated to the post?' ),
        'route'     => JevConnector\Question::choice(
            'Where should this go?',
            array(
                'approve' => 'Clearly a real, on-topic comment',
                'hold'    => 'Plausible but needs a human look',
                'trash'   => 'Abuse or harassment',
            )
        ),
    )
);

if ( ! is_wp_error( $answer ) && $answer->is_confident( 'route', 0.9 ) ) {
    do_something( $answer->choice( 'route' ) );
}

External services

This plugin connects to the TypeSafe System One API, a third-party service operated by TypeSafe, to evaluate the content you pass to it. You supply your own API key.

What is sent: the state you supply to JevConnector\ask() or to the REST route (which may include post content, comment text, or any other data your code passes), the questions you define, and your model identifier. Your API key is sent as a bearer token, and the request’s User-Agent header identifies this plugin, its version, and your site’s home URL.

When it is sent: only when your own code calls JevConnector\ask() or JevConnector\models(), or a permitted user calls one of the REST routes. models() asks only for the list of model names your account can use, sends no content, and is cached for a day. This plugin itself never initiates a request. Nothing is sent until an administrator has connected TypeSafe under Settings → Connectors, and no request is made on activation, on page load, or on any schedule. Identical repeat requests are served from a local cache rather than re-sent.

Where it goes: https://api.typesafe.ai/v1/systemone for questions and https://api.typesafe.ai/v1/models for the model list.

Service terms and policies: https://docs.typesafe.ai/legal
Documentation: https://docs.typesafe.ai