Jev bookmark relevance

Choose which saved reading fits your current task

Provide a reading goal and saved-item summary. Compare useful material for your API-retry work with background reading, unrelated topics and summaries needing review.

Choose a bookmark for this reading goalLive API
Try an example

Original synthetic bookmark metadata; no saved link is opened.

Source: Workflow inspiration on X

Source and example scope

Independent goal-based relevance exercise. The source throughput and comparison claims are not reproduced.

Objective: implement safe API retries in a Node.js service this week.

Decision rules and labels
Read nowSave for laterNot relevantNeeds review
A real response from the Jev API

Edit any input to explore a different decision. No request is sent until you run it.

Your next decision

Not run yet

Choose an example or write your own input, then run it to see the ranked choices.

Results are not stored by this site. Your submitted text is processed by the API provider.

The decision

Relevant to what you need to do next.

This example uses a specific objective: implement safe API retries in a Node.js service this week. A summary that teaches bounded backoff or idempotency fits that task immediately. A general discussion of reliability can still be useful, but it goes into Save for later under this rule. A cake recipe is outside the objective, regardless of its quality.

The labels describe the relationship between a supplied summary and a supplied goal. They do not predict a person's click probability or say whether an article is universally good. Elastic's ranking documentation treats relevance as a search problem, and its evaluation guide recommends representative queries and judgments. Here, six synthetic examples expose the rule rather than estimate performance on a real library.

The source post describes classifying saved bookmarks with Jev. We built a smaller independent exercise with editable summaries, explicit choices and recorded API outputs. The source speed and model-comparison claims are not measurements of this page. The tool reads the text you enter; it does not visit saved URLs or infer the contents of a missing article.

Conceptual saved reading cards separated into a focused reading tray, a later stack and an amber review item.
AI-generated concept illustration of goal-based reading selection; not a real bookmark manager.
How it works

Build a focused reading shortlist.

Keep the goal, summary and result together.

  1. Illustration of adding the text or evidence for a decision

    State the current goal

    Describe a concrete task and what would make a saved item immediately useful.

  2. Illustration of defining a question and its allowed choices

    Supply honest summaries

    Use the title and available summary. Mark absent descriptions as missing instead of guessing the linked content.

  3. Illustration of sending one input and comparing the returned choices

    Inspect the choices

    Compare direct implementation help, related background, unrelated material and insufficient evidence.

  4. Illustration of reviewing a result and an uncertain item before using it

    Review the reading list

    Keep record IDs and the objective. Reclassify when the goal changes and inspect uncertain items before dropping them.

Our API test

Six saved summaries, one explicit reading goal.

Six saved summaries, one explicit reading goal.
Synthetic bookmarkObserved labelLeading probability
Immediate implementationRead now100%
Avoid duplicate writesRead now96%
Related backgroundSave for later97%
Unrelated topicNot relevant100%
No useful summaryNeeds review100%
Related but not immediateSave for later96%

Six completed live Jev requests on September 27, 2026. Earlier attempts encountered provider failures and are excluded from this result table, not counted as successful tests. No accuracy or speed benchmark is claimed.Verified

Understand the results

Try replacing the technical objective with planning a weekend birthday cake

Try replacing the technical objective with planning a weekend birthday cake.

Read the example notes

Rewrite the four label criteria to match that task, then run the cake summary again. The item has not changed, but its usefulness to the current goal has. That is why the reading objective should be saved alongside each classification.

Source and example scope

Needs review handles the vague title with no summary.

Read the example notes

Fetch or write an authorized summary before treating that item as irrelevant. Save for later is different: the supplied description establishes a related topic, but not enough immediate implementation value for this particular week's task.

Using the result in your application

Download observed-bookmarks.json and run-choice-sample.mjs into a private Node.js project.

Read the example notes

Tested setup is Node.js 24.18.0 and @typesafe-ai/sdk 0.6.0. Run npm init -y and npm install @typesafe-ai/sdk@0.6.0, set TYPESAFE_API_KEY privately, then run node run-choice-sample.mjs observed-bookmarks.json sparse. The runner makes one fresh request without retries.

Using the result in your application

Each fixture sample preserves its ID, full summary, policy, choices and recorded output.

Read the example notes

The runner prints the selected ID, elapsed time and provider result. Fresh distributions can differ. If you later build a ranked feed, evaluate it on your own representative goals and judgments; six illustrative checks do not establish reading-list quality or comparative speed.

One live request

Replay one bookmark decision from Node.js.

View the code example
run-choice-sample.mjs
import { readFile } from 'node:fs/promises';
import { choice, TypeSafeClient } from '@typesafe-ai/sdk';

// Run one selected sample. There are no retries or automatic batch requests.
const [file, sampleId] = process.argv.slice(2);
if (!file || !sampleId || !process.env.TYPESAFE_API_KEY) {
  throw new Error('Set TYPESAFE_API_KEY, then run: node run-choice-sample.mjs FIXTURE.json SAMPLE_ID');
}
const raw = await readFile(file, 'utf8');
if (Buffer.byteLength(raw) > 1_000_000) throw new Error('Fixture is too large');
/** @type {{samples: Array<{id: string, request: {context: string, question: string, choices: string[]}}>}} */
const fixture = JSON.parse(raw);
const matches = fixture.samples.filter(sample => sample.id === sampleId);
if (matches.length !== 1) throw new Error('Select exactly one existing sample ID');
const { context, question, choices } = matches[0].request;
if (typeof context !== 'string' || typeof question !== 'string' || context.length > 12000
  || question.length > 500 || !Array.isArray(choices) || choices.length < 2 || choices.length > 12
  || choices.some(label => typeof label !== 'string' || !label || label.length > 120)
  || new Set(choices).size !== choices.length) throw new Error('Invalid sample request');
const client = new TypeSafeClient({ logLevel: 'off' });
const start = performance.now();
try {
  const result = await client.systemOne({ model: 'jev-latest', state: { context, question, choices },
    questions: { decision: choice(question, Object.fromEntries(choices.map(label => [label, null]))) } },
    { timeout: 10000, retry: { maxRetries: 0 } });
  console.log(JSON.stringify({ sampleId, elapsedMs: Math.round(performance.now() - start), result }, null, 2));
} catch {
  console.error('The API request failed. No retry was sent; check your account and network privately.');
  process.exitCode = 1;
}
Scope of this demo

What this example helps you test.

Supplied choices

What you can test

  • Filtering supplied summaries against an explicit reading goal
  • Separating immediate help from related background
  • Retaining sparse summaries for review
Application responsibilities

What this demo does not do

  • Predicting personal engagement from no user evidence
  • Reading unseen URLs or evaluating full article quality
  • Claiming ranking quality from six synthetic examples
Example FAQ

Questions before you use the result.

Understand the choices, the source and what a live result means.

Does the demo open a bookmark URL?

No. It classifies only the title and summary you supply. A missing description remains missing evidence.

Is Read now a measure of article quality?

No. It means the supplied summary directly serves the stated reading goal under this example policy.

Why keep a Save for later option?

Related background may be useful without helping with the immediate implementation task. The extra label preserves that distinction.

Can the same bookmark receive a different label?

Yes. Changing the goal or supplying a better summary changes the evidence used for the decision. Record those inputs with the result.

Does this prove a personalized feed will perform well?

No. These are small synthetic functional checks. A real feed needs representative goals, reviewed relevance judgments and its own evaluation.

Test your own decision

Try another input here or build a different Choice question in the Playground.