Recipe catalog / listing-fact-consistency

Check a listing description against its fact sheet

Does the free-text description contradict the structured facts about the property?

You need to catch a marketing description that disagrees with the structured listing data on bedrooms, size, year built, or features before the listing is published or syndicated.

Explore this recipe interactively · Source and implementation guide

Use listing-fact-consistency in TypeScript

Install with npm install jev-recipes. Requires Node.js 22.9 or newer and ES modules. Set TYPESAFE_API_KEY in your server environment for live calls, which send input to TypeSafe and use API quota. See the installation guide.

import { listingFactConsistency } from 'jev-recipes/listing-fact-consistency';

const result = await listingFactConsistency({
  "description": "Welcome to this beautifully maintained four-bedroom colonial on a half-acre lot. The 2,400 sq ft layout offers a renovated kitchen, hardwood floors throughout, a finished basement, and a two-car attached garage. Built in 1994 and lovingly updated, this home is move-in ready.",
  "facts": "Property type: Single-family, colonial\nBedrooms: 3\nBathrooms: 2.5\nLiving area: 2,380 sq ft\nLot size: 0.5 acres\nYear built: 1994\nGarage: 2-car attached\nBasement: Unfinished\nFlooring: Hardwood, carpet",
  "minConfidence": 0.8
});
console.log(result);

Input contract

{
  "input": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "description": {
        "type": "string"
      },
      "facts": {
        "type": "string"
      },
      "minConfidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      }
    },
    "required": [
      "description",
      "facts"
    ]
  },
  "result": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "model": {
        "type": "string"
      },
      "usage": {
        "type": "object",
        "properties": {
          "input_tokens": {
            "type": "integer",
            "minimum": 0,
            "maximum": 9007199254740991
          },
          "output_tokens": {
            "type": "integer",
            "minimum": 0,
            "maximum": 9007199254740991
          }
        },
        "required": [
          "input_tokens",
          "output_tokens"
        ],
        "additionalProperties": false
      },
      "status": {
        "type": "string",
        "enum": [
          "ready",
          "review"
        ]
      },
      "probability": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      },
      "confidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      },
      "verdict": {
        "type": "string",
        "enum": [
          "contradicts",
          "consistent"
        ]
      }
    },
    "required": [
      "model",
      "usage",
      "status",
      "probability",
      "confidence",
      "verdict"
    ],
    "additionalProperties": false
  }
}

Saved example result

This hand-authored response demonstrates the contract. It is not a model accuracy measurement. Run it without an API key: npx jev-recipes demo listing-fact-consistency.

{
  "model": "demo-fixture",
  "usage": {
    "input_tokens": 0,
    "output_tokens": 0
  },
  "status": "ready",
  "probability": 0.95,
  "confidence": 0.95,
  "verdict": "contradicts"
}

Evaluation evidence

Use the evaluation guide to measure this decision on your own labeled cases.

Limitations

• Reports a single verdict, not which fact is contradicted. Pair it with per-field checks when you need locations.
• Judges only what the description says against what facts states. Neither input is checked against the real property or public records.
• Rounding, unit conversion, and how much tolerance to allow are application decisions; the recipe treats a clear numeric mismatch as a contradiction.

Related recipes

• headline-fit: Use headline-fit to check whether a title represents its body fairly, rather than whether prose contradicts a structured fact sheet.
• extraction-fidelity: Use extraction-fidelity to grade how faithfully structured values were pulled from a source document, the reverse direction of this check.