Recipe catalog / result-plausibility

Check result plausibility

Is result a plausible, internally consistent answer to request rather than an error, placeholder, empty, or unrelated output dressed as data?

You need a yes/no check on a tool or subagent output before an agent trusts it, stores it, or builds the next step on it.

Explore this recipe interactively · Source and implementation guide

Use result-plausibility 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 { resultPlausibility } from 'jev-recipes/result-plausibility';

const result = await resultPlausibility({
  "request": "List the three most recent production deploys with their commit SHAs and deploy times.",
  "result": "[{\"id\": 1, \"sha\": \"abc123\", \"deployed_at\": \"2024-01-01T00:00:00Z\"}, {\"id\": 2, \"sha\": \"abc123\", \"deployed_at\": \"2024-01-01T00:00:00Z\"}, {\"id\": 3, \"sha\": \"abc123\", \"deployed_at\": \"2024-01-01T00:00:00Z\"}]",
  "minConfidence": 0.8
});
console.log(result);

Input contract

{
  "input": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "request": {
        "type": "string"
      },
      "result": {
        "type": "string"
      },
      "minConfidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      }
    },
    "required": [
      "request",
      "result"
    ]
  },
  "result": {
    "$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": [
          "plausible",
          "implausible"
        ]
      }
    },
    "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 result-plausibility.

{
  "model": "demo-fixture",
  "usage": {
    "input_tokens": 0,
    "output_tokens": 0
  },
  "status": "ready",
  "probability": 0.07,
  "confidence": 0.9299999999999999,
  "verdict": "implausible"
}

Evaluation evidence

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

Limitations

• Judges internal consistency and fit to the request, not factual accuracy. A plausible result can still be wrong.
• A genuine empty answer, such as a search that legitimately found nothing, is only plausible if the result says so rather than returning blank output.

Related recipes

• result-usefulness: Use result-usefulness once a result is plausible to decide whether it actually carries information the task needs.
• result-outcome: Use result-outcome to classify what a result reports happened, such as success or failure, rather than whether it is real.