Prompt metadata

Tag dataclass

Tag(value: str, category: str | None = None)

A structured tag attached to a benchmark prompt.

Tags are metadata used to filter and organize benchmark results; they are never shown to the annotators. category optionally groups related tags (e.g. Tag("landscape", category="scene")); leave it None for a bare tag.

Origin dataclass

Origin(source: str)

Where a benchmark prompt originated from (e.g. a source dataset).

BenchmarkPromptInfo dataclass

BenchmarkPromptInfo(
    identifier: str,
    prompt: str | None,
    english_prompt: str | None,
    prompt_asset: list[str] | None,
    tags: list[Tag],
    origin: Origin | None,
)

A single prompt returned by a filtered/sorted benchmark prompt query.

is_tag_list

is_tag_list(value: Any) -> bool

Whether value is usable as one prompt's tag list.

A bare str is itself a Sequence of characters, so it has to be rejected explicitly or tags=["a", "b"] would silently be read as two prompts' worth of tags instead of one.

Source code in src/rapidata/rapidata_client/benchmark/prompt_metadata.py

def is_tag_list(value: Any) -> bool:
    """Whether ``value`` is usable as one prompt's tag list.

    A bare ``str`` is itself a ``Sequence`` of characters, so it has to be
    rejected explicitly or ``tags=["a", "b"]`` would silently be read as two
    prompts' worth of tags instead of one.
    """
    return isinstance(value, Sequence) and not isinstance(value, (str, bytes))

coerce_tags

coerce_tags(tags: Sequence[str | Tag] | None) -> list[Tag]

Normalize user-supplied tags into Tag objects.

Bare strings become uncategorized tags, so callers can keep passing the plain list[str] they used before categories existed.

Source code in src/rapidata/rapidata_client/benchmark/prompt_metadata.py

def coerce_tags(tags: Sequence[str | Tag] | None) -> list[Tag]:
    """Normalize user-supplied tags into ``Tag`` objects.

    Bare strings become uncategorized tags, so callers can keep passing the
    plain ``list[str]`` they used before categories existed.
    """
    return [Tag(value=item) if isinstance(item, str) else item for item in (tags or [])]

coerce_origin

coerce_origin(origin: Origin | str | None) -> Origin | None

Normalize a user-supplied origin into an Origin.

Source code in src/rapidata/rapidata_client/benchmark/prompt_metadata.py

def coerce_origin(origin: Origin | str | None) -> Origin | None:
    """Normalize a user-supplied origin into an ``Origin``."""
    return Origin(source=origin) if isinstance(origin, str) else origin

coerce_prompt_asset

coerce_prompt_asset(asset: object) -> list[str] | None

Normalize one prompt's asset(s) to list[str] | None.

A prompt's assets are a list of URLs / file paths, the same shape the job definitions use for media_contexts; several entries are registered as one multi-asset. A bare str is still accepted for scripts written when a prompt could only carry one asset and is wrapped in a single-element list.

Source code in src/rapidata/rapidata_client/benchmark/prompt_metadata.py

def coerce_prompt_asset(asset: object) -> list[str] | None:
    """Normalize one prompt's asset(s) to ``list[str] | None``.

    A prompt's assets are a list of URLs / file paths, the same shape the job
    definitions use for ``media_contexts``; several entries are registered as
    one multi-asset. A bare ``str`` is still accepted for scripts written when
    a prompt could only carry one asset and is wrapped in a single-element list.
    """
    if asset is None:
        return None
    if isinstance(asset, str):
        if asset == "":
            raise ValueError(
                "A prompt asset cannot be an empty string. If not needed, set to None."
            )
        return [asset]
    if isinstance(asset, list):
        if len(asset) == 0:
            raise ValueError(
                "A prompt asset list cannot be empty. If not needed, set to None."
            )
        if any(not isinstance(item, str) or item == "" for item in asset):
            raise ValueError(
                "Every entry in a prompt asset list must be a non-empty string."
            )
        return asset
    raise ValueError(
        f"A prompt asset must be a list of strings or None, got {type(asset).__name__}."
    )