Recipe catalog / incident-severity-wording

Grade the severity an incident report describes

What severity does the wording of report describe, on a five-level rubric from no user impact to total outage or data loss?

You need a first severity estimate from a free-text incident report, alert summary, or status update before a human incident commander confirms it, or you want to check that a declared severity matches how the report describes the impact.

Explore this recipe interactively · Source and implementation guide

Use incident-severity-wording 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 { incidentSeverityWording } from 'jev-recipes/incident-severity-wording';

const result = await incidentSeverityWording({
  "report": "Since 14:05 UTC, about 30% of requests to the search endpoint on the EU shard return 503. Search results fail to load for mobile users in the EU; web checkout, account pages, and all other regions are unaffected. No data loss observed.",
  "minConfidence": 0.8
});
console.log(result);

Input contract

{
  "input": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "report": {
        "type": "string"
      },
      "minConfidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      }
    },
    "required": [
      "report"
    ]
  },
  "result": {
    "$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"
        ]
      },
      "score": {
        "type": "number",
        "minimum": 0
      },
      "level": {
        "type": "integer",
        "minimum": 0,
        "maximum": 9007199254740991
      },
      "confidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      },
      "probabilities": {
        "type": "object",
        "propertyNames": {
          "type": "string"
        },
        "additionalProperties": {
          "type": "number",
          "minimum": 0,
          "maximum": 1
        }
      },
      "severity": {
        "type": "string",
        "enum": [
          "none",
          "minor",
          "partial",
          "major",
          "critical"
        ]
      }
    },
    "required": [
      "model",
      "usage",
      "status",
      "score",
      "level",
      "confidence",
      "probabilities",
      "severity"
    ],
    "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 incident-severity-wording.

{
  "model": "demo-fixture",
  "usage": {
    "input_tokens": 0,
    "output_tokens": 0
  },
  "status": "ready",
  "score": 2.12,
  "level": 2,
  "confidence": 0.8,
  "probabilities": {
    "0": 0,
    "1": 0.05,
    "2": 0.8,
    "3": 0.13,
    "4": 0.02
  },
  "severity": "partial"
}

Evaluation evidence

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

Limitations

• Grades what the report says, not measured impact. An understated report yields an understated grade.
• Does not know your severity matrix or SLAs. Map levels to your own SEV scale and paging rules in application code.
• Does not distinguish a report written during an incident from one written after recovery.

Related recipes

• issue-impact: Use issue-impact for bug reports and feature requests where the question is how many users a problem affects rather than how large an outage is.
• policy-severity: Use policy-severity to grade how serious a policy violation is, rather than how serious an operational incident is.