Review business contacts against a partnership objective
Supply a business record and the developer-tools partnership rule. Compare relevant current roles with unrelated, outdated or incomplete evidence before maintaining the list.
Check a contact against a working-list objectiveLive API
The decision
How does Jev contact list triage work?
A contact can be useful for one project and irrelevant to another. This example defines a developer-tools partnership objective, then asks Jev to classify only the professional evidence supplied with a record. A current SDK collaboration supports Keep. An explicitly unrelated role supports Exclude from this working list. Missing or contradictory information supports Review.
This is different from qualifying a sales opportunity. A developer community organizer may be relevant to a workshop without having a budget or buying intent. It is also different from duplicate detection. Google Contacts documents a separate process for reviewing and merging duplicate listings; a semantic relevance label does not establish that two records represent the same person.
The X source describes sorting exported contacts with keep, review and removal decisions. We created six independent synthetic records to examine the decision boundary. The example reads the text you provide; it does not connect to LinkedIn, Gmail or a CRM. Source scale claims are not used as measurements of this demo.
AI-generated conceptual illustration of cards routed to working-list groups; not a CRM screenshot.
How it works
Review a working list in four steps.
Make the objective explicit and keep uncertain records visible.
01
Define the objective
Describe which professional roles or collaborations are relevant to this particular list. Keep the rule with the result.
02
Supply current evidence
Use a short, authorized record with relevant professional facts. Leave unknown information unknown and remove unnecessary personal details.
03
Run the three-way decision
Compare clear, unrelated and uncertain records. Inspect alternative probabilities when the result conflicts with your reading.
04
Approve record actions separately
Keep the original ID and inspect Review cases. A relevance label does not merge, delete, enrich or contact a record.
Our API test
Six records, with uncertain roles kept for review.
Six records, with uncertain roles kept for review.
Synthetic contact
Observed label
Leading probability
SDK collaboration
Keep
100%
Developer workshop
Keep
100%
Different field
Exclude
100%
Title only
Review
99%
Old role only
Review
99%
Conflicting fields
Review
94%
Six live Jev requests on September 27, 2026. Synthetic functional checks, not a contact-cleanup accuracy benchmark. Fresh distributions may differ.Verified
Understand the results
Source and example scope
A senior job title is not enough to establish relevance.
Read the example notes
For the title-only example, ask what the organization does and how the person relates to the project. For the old-role example, verify a current role before changing the working list. For conflicting fields, resolve which source is current instead of letting the model silently choose a biography.
Keep the stable record ID with the label and the objective used for the decision
Keep the stable record ID with the label and the objective used for the decision.
Read the example notes
If the partnership objective changes, re-evaluate the records against the new rule. Exclude means outside this particular working list; it is not an instruction to delete a person from the master contact store. The CRM workflow reference likewise distinguishes filtering, merging and archiving as separate operations.
Using the result in your application
Download observed-contacts.json and run-choice-sample.mjs into a private Node.js project.
Read the example notes
We tested Node.js 24.18.0 with @typesafe-ai/sdk 0.6.0. Run npm init -y, then npm install @typesafe-ai/sdk@0.6.0. Set TYPESAFE_API_KEY privately in your shell and run node run-choice-sample.mjs observed-contacts.json vague. The runner sends one fresh API request without retries.
Using the result in your application
The fixture has a samples array with id, request and observed.
Read the example notes
The runner uses context, question and choices from request. It prints the current API response; the original observed answers are retained only for comparison. The six examples are functional checks on invented records, not an estimate of classification accuracy for a real contact database.
Using the result in your application
Try changing the title-only record to a current SDK engineering role while leaving the objective unchanged.
Read the example notes
Then change the objective to restaurant procurement and consider how the relevance policy must change. The exercise makes the dependence on your objective explicit. Approval for outreach, record merges and deletion belongs to separate application controls.
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
Testing objective-relative business relevance
Separating current evidence from vague or stale roles
Building a review queue before changing records
Application responsibilities
What this demo does not do
Inferring buying intent or contact permission
Merging duplicate identities or deleting records
Enriching profiles from external accounts
Example FAQ
Questions before you use the result.
Understand the choices, the source and what a live result means.
Does Exclude delete a contact?
No. It means the supplied record does not fit this example working-list objective. The demo performs no contact-store writes.
Why does an old software role go to Review?
The policy asks for current relevance. An old role alone does not establish what the person does now, so the record needs updated evidence.
Is Keep a qualified sales lead?
No. Relevance to a partnership objective does not establish need, budget, authority, purchase timing or permission to contact someone.
Can the model merge duplicate people?
This example does not assess identity. Use a separate reviewed deduplication workflow and preserve record history when making any merge.
Can I edit the objective?
Yes. Change the policy and input in the hero window. Make the Keep, Review and Exclude criteria consistent with your new objective before interpreting the result.
Test your own decision
Try another input here or build a different Choice question in the Playground.