Skip to content

Translation Task

multilingual_translation(target_languages=["ja"]) requests only Japanese and produces a schema with only the required ja field. Pass multiple supported language codes to request several translations; omitting target_languages preserves the original all-51-language response model (TranslatedString). Empty, duplicate, and unknown codes are rejected.

The Icelandic JSON key is is: access it in Python as result.is_, and use result.model_dump(by_alias=True) to serialize language-code keys. The full model still accepts is_ when validating older Python-side data.

openaivec.task.nlp.translation

Multilingual translation task definition.

Classes

TranslatedString

Bases: BaseModel

Translations for a fixed set of language-code fields.

Functions:

multilingual_translation 

multilingual_translation(
    target_languages: Sequence[str] | None = None,
) -> PreparedTask[BaseModel]

Create a translation task for selected language codes.

Parameters:

Name Type Description Default
target_languages Sequence[str] | None

Language codes to translate into. None (the default) requests all 51 supported languages. Use is for Icelandic; the Python attribute remains is_ for compatibility.

 None

Returns:

Type Description
PreparedTask[BaseModel]

PreparedTask[BaseModel]: Task with exactly the requested response fields.

Raises:

Type Description
TypeError

If target_languages is a string rather than a sequence.
ValueError

If the selection is empty, duplicated, or unsupported.
Source code in src/openaivec/task/nlp/translation.py 
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
def multilingual_translation(target_languages: Sequence[str] | None = None) -> PreparedTask[BaseModel]:
    """Create a translation task for selected language codes.

    Args:
        target_languages (Sequence[str] | None): Language codes to translate into.
            ``None`` (the default) requests all 51 supported languages. Use ``is``
            for Icelandic; the Python attribute remains ``is_`` for compatibility.

    Returns:
        PreparedTask[BaseModel]: Task with exactly the requested response fields.

    Raises:
        TypeError: If target_languages is a string rather than a sequence.
        ValueError: If the selection is empty, duplicated, or unsupported.
    """
    supported = {field.alias or name: name for name, field in TranslatedString.model_fields.items()}
    if isinstance(target_languages, str):
        raise TypeError("target_languages must be a sequence of language codes, not a string")
    codes = tuple(supported) if target_languages is None else tuple(target_languages)
    if (
        not codes
        or any(not isinstance(code, str) or code not in supported for code in codes)
        or len(codes) != len(set(codes))
    ):
        raise ValueError("target_languages must contain unique supported language codes")
    response_format = TranslatedString
    if target_languages is not None:
        fields: dict[str, Any] = {
            supported[code]: (str, TranslatedString.model_fields[supported[code]]) for code in codes
        }
        response_format = create_model(
            "SelectedTranslations",
            __config__=ConfigDict(extra="forbid", populate_by_name=True),
            __module__=__name__,
            **fields,
        )
    return PreparedTask(
        instructions=_build_instructions(codes),
        response_format=response_format,
    )