Recipe catalog / change-window-fit

Check a change against a change-window policy

Does the described change fall within the written change-window or freeze policy?

A deploy or infrastructure change is proposed in free text and you want a first read on whether it is permitted under a written change-window or freeze policy before a human approver looks at it.

Explore this recipe interactively · Source and implementation guide

Use change-window-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 { changeWindowFit } from 'jev-recipes/change-window-fit';

const result = await changeWindowFit({
  "change": "Deploy payments-service 4.12.1 to production on Friday 2026-09-25 at 16:30 UTC. It patches a rounding bug in refund amounts. Standard rolling deploy, no schema change. Not an emergency; the bug affects about 0.1% of refunds.",
  "policy": "Production change windows are Monday to Thursday, 09:00 to 17:00 UTC. No production deploys on Fridays, weekends, or during the quarter-end freeze covering the last five business days of each quarter. Emergency fixes may proceed outside windows only with written approval from the incident commander.",
  "minConfidence": 0.8
});
console.log(result);

Input contract

{
  "input": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "change": {
        "type": "string"
      },
      "policy": {
        "type": "string"
      },
      "minConfidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      }
    },
    "required": [
      "change",
      "policy"
    ]
  },
  "result": {
    "$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": [
          "allowed",
          "blocked"
        ]
      }
    },
    "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 change-window-fit.

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

Evaluation evidence

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

Limitations

• Dates, times, time zones, and business-day arithmetic must be compared in application code. The recipe reads the texts as written and does not compute calendar facts.
• Judges the change as described against the policy as written. It does not know about approvals, exceptions, or freeze calendars that are not in policy.
• A blocked verdict is a flag for an approver, not an enforcement decision.

Related recipes

• slot-fit: Use slot-fit to check whether a proposed time matches a stated availability window in scheduling contexts, rather than a deployment policy.
• action-scope: Use action-scope to check whether an action stays within what was requested, rather than whether its timing is allowed.