Jev Citation Verifier

01 / LIVE VERIFIER

Does this source actually support this claim?

The supported example is already filled in. Click Verify citation without editing anything. The quote check runs in your browser; only a verified quote reaches the real TypeSafe Jev model.

02 / HOW IT WORKS

One workflow, three different responsibilities

“Supports” is not the same as “true.” Jev judges the relationship between the supplied claim and supplied evidence. A source can support a claim and still be unreliable, outdated, or incomplete.

03 / WHY VERIFY THE QUOTE FIRST

Use each tool for the job it does well

04 / HOW TO READ THE RESULT

Relation, probability, then review status

The selected relation comes from Jev. Its actual probability drives the local demo policy: at or above the chosen threshold the status is Auto; otherwise it is Human Review. A missing distribution always goes to Human Review. Moving the threshold never calls the model again.

05 / REAL-WORLD PATTERN

The full system can start much earlier

The public Paper Trellis Citation Verifier demonstrates a broader workflow: parse a manuscript, match references, locate source text, verify proposed quotes in code, use Jev for semantic support, and leave the final decision to a person. This page demonstrates only the core verification layer after the evidence has been supplied.

Attributed example: Paper Trellis Citation Verifier on GitHub ↗ (opens in a new tab).

06 / IMPLEMENTATION

The same two-stage check in TypeScript

This original example uses the same Vercel AI Gateway integration as Jev Playground. It stops before the model when the quote is absent, asks one Choice question when it is present, and applies a local review threshold afterward.

import { createGateway, experimental_evaluate as evaluate } from 'ai';

const gateway = createGateway({
  apiKey: process.env.AI_GATEWAY_API_KEY,
});

const normalizeLines = (text: string) => text.replace(/\r\n?/g, '\n');

export async function verifyCitation(
  claim: string,
  quote: string,
  source: string,
) {
  const quoteFound = normalizeLines(source).includes(normalizeLines(quote));

  if (!quoteFound) throw new Error('Exact quote not found');

  const result = await evaluate({
    model: gateway.evaluationModel('typesafe-ai/jev'),
    state: `CLAIM:
${claim}

VERIFIED QUOTE:
${quote}

SOURCE:
${source}`,
    questions: {
      relation: {
        type: 'choice',
        instructions: 'How does the verified evidence relate to the claim?',
        criteria: {
          SUPPORTS: 'The evidence directly supports or clearly entails the claim.',
          CONTRADICTS: 'The evidence states or clearly implies the opposite.',
          INSUFFICIENT: 'Relevant, but not enough to establish the claim.',
          UNRELATED: 'The evidence does not address the claim.',
        },
      },
    },
    maxRetries: 0,
    abortSignal: AbortSignal.timeout(30_000),
    providerOptions: { gateway: { zeroDataRetention: true } },
  });

  const answer = result.answers.relation;
  const topProbability = answer.probabilities?.[answer.choice];
  const reviewThreshold = 0.8; // Your application policy, not Jev's.
  const status = topProbability !== undefined && topProbability >= reviewThreshold
    ? 'AUTO'
    : 'HUMAN REVIEW';

  return { relation: answer.choice, probabilities: answer.probabilities, status };
}

For request contracts, production error handling, and server-side credentials, continue to the Jev API guide.

07 / LIMITATIONS

What this verification cannot prove

• The source itself is trustworthy.
• The source is current.
• The claim is universally true.
• The quote is interpreted correctly across the full document.
• An academic citation meets a journal’s standards.

The demo evaluates only the relationship between the supplied claim and supplied evidence.