prompts.chatprompts.chat
Login
All classifiers
ScoreIT & incidents

Proposed incident severity

Propose a severity level from the incident description. The incident commander confirms it.

F@f

Limits to keep in mind

A proposal only. The incident commander sets the severity, and it can change as more is learned.

Problem

Severity decides who gets paged and how fast communication goes out, but the first responder often picks it under pressure from incomplete information.

Goal

Give the responder a consistent starting proposal and page the incident commander quickly for the highest levels.

Input / state

What's known so far: the alert, customer impact, and the responder's first notes.

summary
What is failing and since when.
customer_impact
Customer impact signal and any numbers known.
notes
The responder's first notes.

Decision question

ScoreQuestion key severity

Based on what is known, which severity level fits this incident best?

Rubric levels

  • 0Minor: no customer impact, or a cosmetic problem.
  • 1Degraded: a feature is slow or failing for some users, and a workaround exists.
  • 2Major: a core feature is failing for many users with no workaround.
  • 3Critical: a full outage, risk of data loss, or a security exposure.

Allowed outcomes

Decided in your code from the answer and its confidence. Thresholds are yours to set and test.

  • Propose minorMost likely level is 0 and confidence meets your threshold.
  • Propose degradedMost likely level is 1 and confidence meets your threshold.
  • Propose majorMost likely level is 2 and confidence meets your threshold.
  • Propose criticalMost likely level is 3 and confidence meets your threshold.
  • Responder decidesConfidence is below your threshold: probability is spread across options.

Example scenarios

Predefined demonstrations. No model is called; the outcome shown is an illustrative judgment.

Example state

{
  "notes": "cosmetic",
  "summary": "Footer link to the careers page 404s",
  "customer_impact": "none reported"
}

Outcome

Illustrative
Propose minor

Most likely level is 0 and confidence meets your threshold.

Why

Cosmetic, no impact.

Using it well

  • Replace the level descriptions with your own severity definitions; the question only works as well as the rubric.
  • When probability is split between two high levels, page as if it were the higher one and let the commander downgrade.

Request definition

POST https://api.typesafe.ai/v1/systemone
{
  "model": "jev-latest",
  "state": {
    "summary": "<summary>",
    "customer_impact": "<customer_impact>",
    "notes": "<notes>"
  },
  "questions": {
    "severity": {
      "type": "score",
      "instructions": "Based on what is known, which severity level fits this incident best?",
      "criteria": [
        "Minor: no customer impact, or a cosmetic problem.",
        "Degraded: a feature is slow or failing for some users, and a workaround exists.",
        "Major: a core feature is failing for many users with no workaround.",
        "Critical: a full outage, risk of data loss, or a security exposure."
      ]
    }
  }
}

Request shape from the System One API reference; a Kev server accepts the same body. Replace the placeholder state with your data. Requests need your own API key or server.

Version History

v1Current Versionabout 8 hours ago@f

Initial version

Comments (0)