Recipe catalog / repetition-level

Grade how repetitive recent material is

How repetitive is recentMaterial, from varied to stuck on one idea?

A generation loop such as an infinite piano player should notice when it has fallen into a loop or frozen on one figure, so it can inject variation or change direction.

Explore this recipe interactively · Source and implementation guide

Use repetition-level 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 { repetitionLevel } from 'jev-recipes/repetition-level';

const result = await repetitionLevel({
  "recentMaterial": "C4 E4 G4 E4 | C4 E4 G4 E4 | C4 E4 G4 E4 | C4 E4 G4 E4 | C4 E4 G4 E4 | C4 E4 G4 E4 (all quarter notes, same dynamic, no rests)",
  "minConfidence": 0.8
});
console.log(result);

Input contract

{
  "input": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "recentMaterial": {
        "type": "string"
      },
      "minConfidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      }
    },
    "required": [
      "recentMaterial"
    ]
  },
  "result": {
    "$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
        }
      },
      "repetition": {
        "type": "string",
        "enum": [
          "varied",
          "some",
          "repetitive",
          "looping",
          "stuck"
        ]
      }
    },
    "required": [
      "model",
      "usage",
      "status",
      "score",
      "level",
      "confidence",
      "probabilities",
      "repetition"
    ],
    "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 repetition-level.

{
  "model": "demo-fixture",
  "usage": {
    "input_tokens": 0,
    "output_tokens": 0
  },
  "status": "ready",
  "score": 2.94,
  "level": 3,
  "confidence": 0.85,
  "probabilities": {
    "0": 0.01,
    "1": 0.01,
    "2": 0.07,
    "3": 0.85,
    "4": 0.06
  },
  "repetition": "looping"
}

Evaluation evidence

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

Limitations

• Grades repetition as written in the note names, rhythms, and annotations. It does not detect transpositions, inversions, or rhythmic augmentations exactly; describe them in the text when they matter.
• Repetition is not a fault in itself. Ostinatos, riffs, and minimalism repeat by design, so the threshold for intervening belongs in application code.
• Only the supplied material is graded; earlier material the caller left out does not count.

Related recipes

• repeated-attempt: Use repeated-attempt to check whether one proposed attempt repeats a previous approach to a task, rather than grading the repetition across a stretch of music.
• progress-stall: Use progress-stall to detect an agent circling on a task objective, not a melody circling on one figure.