Walk a label tree | milliseconds.ai documentation

curl -X POST https://api.milliseconds.ai/v1/decision-machine-1/classify-tree \
     -H "Authorization: Bearer <token>" \
     -H "Content-Type: application/json" \
     -d '{
  "tree": {
    "cardiovascular": {
      "description": "heart and blood vessels: chest pain, arrhythmia, heart failure, hypertension",
      "labels": {
        "acute_coronary_syndrome": {
          "description": "heart attack or angina: chest pain, ST elevation on the ECG, troponin rise",
          "labels": {
            "nstemi": "non-ST-elevation myocardial infarction: troponin rise without ST elevation",
            "stemi": "ST-elevation myocardial infarction: ST elevation on the ECG",
            "unstable_angina": "chest pain at rest without a troponin rise"
          }
        },
        "arrhythmia": "abnormal heart rhythm: palpitations, atrial fibrillation, bradycardia",
        "heart_failure": "fluid overload, breathlessness on exertion, swollen legs, reduced ejection fraction"
      }
    },
    "gastrointestinal": "digestive tract: abdominal pain, nausea, vomiting, bleeding",
    "neurological": "brain and nerves: headache, seizure, weakness, numbness",
    "pulmonary": {
      "description": "lungs and airways: cough, breathlessness, wheeze, pneumonia",
      "labels": {
        "embolism": "pulmonary embolism: sudden breathlessness, pleuritic pain, low oxygen",
        "infection": "pneumonia or bronchitis: fever with a productive cough",
        "obstructive": "asthma or COPD: wheeze, chronic cough, smoking history"
      }
    }
  },
  "text": "62-year-old man with crushing substernal chest pain radiating to the left arm for 40 minutes, sweating, and ST elevation in leads II, III and aVF."
}'

{
  "confidence": 0.524,
  "label": "stemi",
  "levels": [
    {
      "confidence": 0.931,
      "inference_ms": 57,
      "input_chars": 456,
      "input_tokens": 112,
      "label": "cardiovascular",
      "probability": 0.982,
      "scores": {
        "cardiovascular": 0.982,
        "gastrointestinal": 0.001,
        "neurological": 0.016,
        "pulmonary": 0.001
      }
    },
    {
      "confidence": 0.564,
      "inference_ms": 56,
      "input_chars": 428,
      "input_tokens": 105,
      "label": "acute_coronary_syndrome",
      "probability": 0.825,
      "scores": {
        "acute_coronary_syndrome": 0.825,
        "arrhythmia": 0.173,
        "heart_failure": 0.003
      }
    },
    {
      "confidence": 0.998,
      "inference_ms": 59,
      "input_chars": 357,
      "input_tokens": 88,
      "label": "stemi",
      "probability": 1,
      "scores": {
        "nstemi": 0,
        "stemi": 1,
        "unstable_angina": 0
      }
    }
  ],
  "path": [
    "cardiovascular",
    "acute_coronary_syndrome",
    "stemi"
  ],
  "probability": 0.81
}

Runs classify once per level of a nested label tree, on the same text, descending into the winning label each time. Returns the path, the deepest label, the compound probability and confidence (products over the levels), and one entry per level with its scores, the input it sent to the model and its inference time.

Send text for one result, texts for a batch wrapped as { results: [...] }, or image (one image, at most 5 MB decoded, optionally with text; never with texts). The response carries x-input-chars (input characters) and x-input-tokens (input tokens billed for this call) headers. Each item of texts (and each of statements or questions) counts as one request against the requests-per-minute limit. Each level re-sends the text, so latency grows with depth.

Authentication

AuthorizationBearer

A milliseconds API key, test_sk-{namespace}-{entropy} or prod_sk-{namespace}-{entropy}. Every /v1 route requires it.

Request

This endpoint expects an object.

treestring or objectRequired

textstringOptional1-20000 characters

The text to decide about.

textslist of stringsOptional

Batch form: one result per text.

imagestringOptional>=1 character

One image: a data:image/(jpeg|png|webp);base64, URL, or bare base64 of a JPEG, PNG or WebP. 5 MB decoded at most. URLs are never fetched. May travel with text (extra context), never with texts.

detailenumOptional

Image resolution: low 512 px, medium 768 px (default), high 1024 px longest edge. Higher sees more and bills more.

Allowed values:

questionstringOptional1-2000 characters

The question the model answers by choosing one label or level. Defaults to a generic one; a real rubric (what to look for, how to treat missing evidence) improves hard cases.

Response

Result, or { results } for a batch

ClassifyTreeResultobject

object

Errors

400

Bad Request Error

502

Bad Gateway Error