Multiple questions

Ask up to 16 questions about the same context in one request: one HTTP call, one model job, one answer per question.

A single request can ask several questions about the same context. Krun answers all of them together:

1 HTTP request  →  1 model job  →  N answers

This is faster and simpler than one request per question: you make one round trip, handle one response and one request id, and all answers come from the same model job.

Example

Classify a support message by department, priority and sentiment in one call:

Python

from krun import Krun

client = Krun()

result = client.decide(
    context="My package never arrived and I need it for a wedding this Saturday.",
    questions={
        "department": {
            "type": "choice",
            "options": {
                "shipping": "Shipping and delivery issues",
                "returns": "Returns and refunds",
                "billing": "Billing and payment issues",
            },
        },
        "priority": {
            "type": "choice",
            "options": {
                "low": "Can wait",
                "normal": "Normal priority",
                "high": "Needs quick attention",
            },
        },
        "sentiment": {
            "type": "choice",
            "options": {"positive": "", "neutral": "", "negative": ""},
        },
    },
)

for question_id, answer in result.answers.items():
    print(question_id, answer.choice, answer.abstain)

Response

{
  "model": "krun-one-v0",
  "answers": {
    "department": {
      "type": "choice",
      "choice": "shipping",
      "confidence": 0.964244,
      "probabilities": { "shipping": 0.978313, "returns": 0.01407, "billing": 0.007617 },
      "abstain": false,
      "abstention_status": "advisory"
    },
    "priority": {
      "type": "choice",
      "choice": null,
      "confidence": 0.388209,
      "probabilities": { "low": 0.228391, "normal": 0.6166, "high": 0.155009 },
      "abstain": true,
      "abstention_status": "advisory"
    },
    "sentiment": {
      "type": "choice",
      "choice": "negative",
      "confidence": 0.998928,
      "probabilities": { "positive": 0.00009, "neutral": 0.000491, "negative": 0.999419 },
      "abstain": false,
      "abstention_status": "calibrated"
    }
  },
  "usage": { "input_tokens": 147 }
}

How answers work

• Same ids, same order. answers has one entry per question, under your question ids, in request order.
• Independent decisions. Each question is scored on its own. One answer can abstain while the others don’t, like priority above.
• Per-answer abstention status. Each answer has its own abstention_status. Here sentiment uses label-only options, so it is calibrated. The others have descriptions, so they are advisory.
• Summed usage. usage.input_tokens is the sum over all questions. See Usage and tokens.
• One request id for the whole request. Feedback names the question with question_id. See Feedback.

Limits

Limit	Value
Questions per request	1 to 16
Options per question	2 to 64
Question id	1 to 100 characters, unique in the request

If any question is invalid, the whole request fails with 400 and no question is answered.

When to use separate requests

Use separate requests when the questions are about different contexts, or when the second question depends on the answer to the first, such as choosing an intent inside a category.