Модели /

Respan: Span-01 Lite (free)

Идентификатор модели

respan/span-01-lite:free

Характеристики

Контекстное окно
—
Архитектура
text->decisions

Модальности

Входящие

  • text

Исходящие

  • decisions

Цены

Оплата по факту за использованные токены и операции. Цены в рублях (за 1 млн токенов или за единицу — как указано).

Тип Стоимость
Входящие токены 0.00 ₽ / 1M

Цены могут меняться в зависимости от курса валют. Данные по тарифам поступают от поставщика услуг и обновляются каждые сутки.

Описание

Here's a thinking process:

  1. Analyze User Request:
  • Task: Translate a model description from English to Russian
  • Source Text: "Модель: Respan: Span-01 Lite (free) (respan/span-01-lite:free)" followed by the original description "Span-01 Lite is the free, lighter tier of Span-01, a behavior scoring model from Respan. It returns, for each plain-language behavior you define, the probability that the behavior is present..."
  • Constraints:
    • Keep model names and technical terms in the familiar format for Russian-speaking developers
    • Output only the text translation, no introduction, no quotes around the whole text
  • Input Format: The user provided a mix of catalog header and description. I need to translate the description part, keeping model names/terms as specified.
  1. Identify Key Elements to Preserve:
  • "Respan" - likely a company/project name, keep as is
  • "Span-01 Lite" - model name, keep as is (familiar format for devs)
  • "(free)" - keep as is, or maybe translate "free" if it's part of the name, but instructions say keep technical terms in familiar format. I'll keep "(free)" as is since it's part of the catalog entry.
  • "respan/span-01-lite:free" - technical repo/path, keep as is
  • "behavior scoring model" - technical term, keep familiar format. In Russian dev context, this might be "модель оценки поведения" or keep "behavior scoring model" but the instruction says "keep names and technical terms in familiar for Russian-speaking developers format". I'll translate the descriptive part but keep the term "behavior scoring" as is or adapt it naturally. Actually, re-read: "Сохраняй имена моделей и технические термины в привычном для русскоязычных разработчиков виде." This means I should keep model names exactly as they are (Span-01 Lite, respan, etc.), and for technical terms, use the standard Russian dev jargon or keep them in English if that's the norm. Often, Russian devs keep English terms like "model", "probability", "behavior", but I can translate the surrounding description. Let me parse carefully.
  • The text: "Span-01 Lite is the free, lighter tier of Span-01, a behavior scoring model from Respan. It returns, for each plain-language behavior you define, the probability that the behavior is present..."
  • I need to translate this into Russian, but keep "Span-01 Lite", "Span-01", "Respan", "behavior scoring model", "plain-language behavior", "probability" etc. in a format familiar to Russian devs. Usually, this means either keeping them in English or transliterating/translating appropriately. Given the instruction "сохраняй имена моделей и технические термины в привычном для русскоязычных разработчиков виде", I think it means: don't translate the model names (keep Span-01 Lite, respan/span-01-lite:free), and for terms like "behavior scoring model", I can either keep in English or use the Russian equivalent that devs use. I'll keep the technical terms in English as they are standard, but the sentence structure in Russian. Actually, let's look at the exact phrasing: "Span-01 Lite is the free, lighter tier of Span-01, a behavior scoring model from Respan." I'll translate "is the free, lighter tier of" but keep the model names. Maybe "Span-01 Lite — это бесплатная, легкая версия Span-01, модель оценки поведения от Respan." But "behavior scoring model" might be better as "модель оценки поведения" or kept in English. The instruction says "technical terms in familiar format for Russian-speaking developers". In RuDev, it's common to see "модель scoring" or just keep English. I'll use Russian translation for the descriptive parts, and keep the key terms as is, maybe in parentheses or naturally integrated. To be safe, I'll keep "behavior scoring model" as "behavior scoring model" (since it's a technical term), or translate it to "модель оценки поведения". Let's see what's more "familiar for Russian-speaking developers". Often, they'd say "модель оценки поведения". I'll go with translating the meaning but keeping the model names exactly.

Let me re-read carefully: "Сохраняй имена моделей и технические термины в привычном для русскоязычных разработчиков виде." This likely means: keep model names like "Span-01 Lite", "respan/span-01-lite:free" exactly as they are. For technical terms, use the standard Russian dev terminology, which often means keeping English terms if they're standard, or using the Russian equivalent that's widely accepted. I think it's safest to keep the English technical terms as is (since they're proper nouns/terms), and translate the rest into Russian sentence structure.

Original: "Span-01 Lite is the free, lighter tier of Span-01, a behavior scoring model from Respan. It returns, for each plain-language behavior you define, the probability that the behavior is present..."

Translation plan:

  • "Span-01 Lite" stays
  • "is the free, lighter tier of" -> "является бесплатной, более легкой версией" or "является…

Решения

Модель не пишет текст. В state передаются данные, в questions — вопросы с известными вариантами. Ответ приходит готовыми значениями.

  • noul — да или нет, вероятность от 0 до 1
  • choice — один вариант из объекта criteria
  • score — положение на шкале, уровни задаются массивом criteria
import requests

response = requests.post(
    "https://routerapi.ru/api/v1/decisions",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={
        "model": "respan/span-01-lite:free",
        "state": "С 10:40 все запросы к API возвращают 500.",
        "questions": {
            "is_urgent": {
                "type": "noul",
                "instructions": "Требует ли обращение немедленной реакции?",
                "criteria": {"true": "Сервис недоступен", "false": "Можно разобрать в очереди"},
            },
            "team": {
                "type": "choice",
                "instructions": "На какую команду направить обращение?",
                "criteria": {"billing": "Оплата и счета", "technical": "Ошибки API", "other": "Всё остальное"},
            },
            "severity": {
                "type": "score",
                "instructions": "Насколько сильно это влияет на работу клиента?",
                "criteria": ["Влияния нет", "Есть обходной путь", "Работа заблокирована"],
            },
        },
    },
)
answers = response.json()["answers"]
print(answers["is_urgent"]["noul"], answers["team"]["choice"], answers["severity"]["score"])

Ответы приходят в answers под теми же ключами, что и вопросы.