Recipe catalog / corrective-action-fit

Check a corrective action against its root cause

Does action address the cause stated in rootCause rather than only the symptom or the affected items?

You review corrective action plans in a CAPA or nonconformance workflow and want to flag actions that rework parts, add inspection, or retrain people while leaving the stated cause in place.

Explore this recipe interactively · Source and implementation guide

Use corrective-action-fit 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 { correctiveActionFit } from 'jev-recipes/corrective-action-fit';

const result = await correctiveActionFit({
  "rootCause": "The torque wrench at station 4 was reading 15% low because the tooling register lists a 24-month calibration interval for this wrench model while the manufacturer specifies 12 months. The wrench passed its last register-scheduled check and drifted out of tolerance eight months later, so every assembly torqued at station 4 since then is suspect.",
  "action": "Rework the 340 assemblies built since the last calibration to the specified torque and verify with a calibrated reference wrench. Send the station 4 wrench for recalibration and return it to service once it passes. Brief operators on checking the calibration sticker before each shift.",
  "minConfidence": 0.8
});
console.log(result);

Input contract

{
  "input": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "rootCause": {
        "type": "string"
      },
      "action": {
        "type": "string"
      },
      "minConfidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      }
    },
    "required": [
      "rootCause",
      "action"
    ]
  },
  "result": {
    "$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": [
          "addresses",
          "misses"
        ]
      }
    },
    "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 corrective-action-fit.

{
  "model": "demo-fixture",
  "usage": {
    "input_tokens": 0,
    "output_tokens": 0
  },
  "status": "ready",
  "probability": 0.08,
  "confidence": 0.92,
  "verdict": "misses"
}

Evaluation evidence

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

Limitations

• Judges the fit between two texts as written. It does not know whether rootCause is the true cause or whether action will work in practice.
• An action can address the cause and still be incomplete, unowned, or late. Owners, due dates, and effectiveness checks belong in application code.
• A misses verdict is a flag for a reviewer, not a rejection.

Related recipes

• workaround-fit: Use workaround-fit to check whether a temporary workaround addresses an issue within stated constraints, rather than whether a permanent action removes a cause.
• troubleshooting-fit: Use troubleshooting-fit to assess a diagnostic procedure rather than a proposed fix.