mirascope.core.base.prompt

The BasePrompt class for better prompt engineering.

BasePrompt

Bases: BaseModel

The base class for engineering prompts.

This class is implemented as the base for all prompting needs. It is intended to work across various providers by providing a common prompt interface.

Example:

from mirascope.core import BasePrompt, metadata, prompt_template

@prompt_template("Recommend a {genre} book")
@metadata({"tags": {"version:0001", "books"}})
class BookRecommendationPrompt(BasePrompt):
    genre: str

prompt = BookRecommendationPrompt(genre="fantasy")

print(prompt)
# > Recommend a fantasy book

print(prompt.message_params())
# > [BaseMessageParam(role="user", content="Recommend a fantasy book")]

print(prompt.dump()["metadata"])
# > {"metadata": {"version:0001", "books"}}

message_params

message_params() -> list[BaseMessageParam]

Returns the list of parsed message parameters.

Source code in mirascope/core/base/prompt.py

def message_params(self) -> list[BaseMessageParam]:
    """Returns the list of parsed message parameters."""
    return parse_prompt_messages(
        roles=SUPPORTED_MESSAGE_ROLES,
        template=get_prompt_template(self),
        attrs={field: getattr(self, field) for field in self.model_fields},
    )

dynamic_config

dynamic_config() -> BaseDynamicConfig

Returns the dynamic config of the prompt.

Source code in mirascope/core/base/prompt.py

def dynamic_config(self) -> BaseDynamicConfig:
    """Returns the dynamic config of the prompt."""
    return None

dump

dump() -> dict[str, Any]

Dumps the contents of the prompt into a dictionary.

Source code in mirascope/core/base/prompt.py

def dump(self) -> dict[str, Any]:
    """Dumps the contents of the prompt into a dictionary."""
    return {
        "metadata": get_metadata(self, None),
        "prompt": str(self),
        "template": get_prompt_template(self),
        "inputs": self.model_dump(),
    }

run

run(
    call_decorator: (
        Callable[
            [Callable[..., BaseDynamicConfig]],
            Callable[..., _BaseCallResponseT],
        ]
        | Callable[
            [Callable[..., BaseDynamicConfig]],
            Callable[..., _BaseStreamT],
        ]
        | Callable[
            [Callable[..., BaseDynamicConfig]],
            Callable[..., _ResponseModelT],
        ]
        | Callable[
            [Callable[..., BaseDynamicConfig]],
            Callable[..., Iterable[_ResponseModelT]],
        ]
    ),
    *additional_decorators: Callable[[_T], _T]
) -> (
    _BaseCallResponseT
    | _BaseStreamT
    | _ResponseModelT
    | Iterable[_ResponseModelT]
)

Returns the response of calling the API of the provided decorator.

Example:

from mirascope.core import BasePrompt, openai, prompt_template


@prompt_template("Recommend a {genre} book")
class BookRecommendationPrompt(BasePrompt):
    genre: str


prompt = BookRecommendationPrompt(genre="fantasy")
response = prompt.run(openai.call("gpt-4o-mini"))
print(response.content)

Source code in mirascope/core/base/prompt.py

def run(
    self,
    call_decorator: Callable[
        [Callable[..., BaseDynamicConfig]],
        Callable[..., _BaseCallResponseT],
    ]
    | Callable[
        [Callable[..., BaseDynamicConfig]],
        Callable[..., _BaseStreamT],
    ]
    | Callable[
        [Callable[..., BaseDynamicConfig]],
        Callable[..., _ResponseModelT],
    ]
    | Callable[
        [Callable[..., BaseDynamicConfig]],
        Callable[..., Iterable[_ResponseModelT]],
    ],
    *additional_decorators: Callable[[_T], _T],
) -> (
    _BaseCallResponseT | _BaseStreamT | _ResponseModelT | Iterable[_ResponseModelT]
):
    """Returns the response of calling the API of the provided decorator.

    Example:

    ```python
    from mirascope.core import BasePrompt, openai, prompt_template


    @prompt_template("Recommend a {genre} book")
    class BookRecommendationPrompt(BasePrompt):
        genre: str


    prompt = BookRecommendationPrompt(genre="fantasy")
    response = prompt.run(openai.call("gpt-4o-mini"))
    print(response.content)
    ```
    """
    kwargs = {field: getattr(self, field) for field in self.model_fields}
    args_str = ", ".join(kwargs.keys())
    namespace, fn_name = {}, self.__class__.__name__
    exec(f"def {fn_name}({args_str}): ...", namespace)
    return reduce(
        lambda res, f: f(res),  # pyright: ignore [reportArgumentType, reportCallIssue]
        [
            metadata(get_metadata(self, self.dynamic_config())),
            prompt_template(get_prompt_template(self)),
            call_decorator,
            *additional_decorators,
        ],
        namespace[fn_name],
    )(**kwargs)

run_async

run_async(
    call_decorator: (
        Callable[
            [Callable[..., Awaitable[BaseDynamicConfig]]],
            Callable[..., Awaitable[_BaseCallResponseT]],
        ]
        | Callable[
            [Callable[..., Awaitable[BaseDynamicConfig]]],
            Callable[..., Awaitable[_BaseStreamT]],
        ]
        | Callable[
            [Callable[..., Awaitable[BaseDynamicConfig]]],
            Callable[..., Awaitable[_ResponseModelT]],
        ]
        | Callable[
            [Callable[..., Awaitable[BaseDynamicConfig]]],
            Callable[
                ...,
                Awaitable[AsyncIterable[_ResponseModelT]],
            ],
        ]
    ),
    *additional_decorators: Callable[[_T], _T]
) -> (
    Awaitable[_BaseCallResponseT]
    | Awaitable[_BaseStreamT]
    | Awaitable[_ResponseModelT]
    | Awaitable[AsyncIterable[_ResponseModelT]]
)

Returns the response of calling the API of the provided decorator.

Example:

import asyncio

from mirascope.core import BasePrompt, openai, prompt_template


@prompt_template("Recommend a {genre} book")
class BookRecommendationPrompt(BasePrompt):
    genre: str


async def run():
    prompt = BookRecommendationPrompt(genre="fantasy")
    response = await prompt.run_async(openai.call("gpt-4o-mini"))
    print(response.content)


asyncio.run(run())

Source code in mirascope/core/base/prompt.py

def run_async(
    self,
    call_decorator: Callable[
        [Callable[..., Awaitable[BaseDynamicConfig]]],
        Callable[..., Awaitable[_BaseCallResponseT]],
    ]
    | Callable[
        [Callable[..., Awaitable[BaseDynamicConfig]]],
        Callable[..., Awaitable[_BaseStreamT]],
    ]
    | Callable[
        [Callable[..., Awaitable[BaseDynamicConfig]]],
        Callable[..., Awaitable[_ResponseModelT]],
    ]
    | Callable[
        [Callable[..., Awaitable[BaseDynamicConfig]]],
        Callable[..., Awaitable[AsyncIterable[_ResponseModelT]]],
    ],
    *additional_decorators: Callable[[_T], _T],
) -> (
    Awaitable[_BaseCallResponseT]
    | Awaitable[_BaseStreamT]
    | Awaitable[_ResponseModelT]
    | Awaitable[AsyncIterable[_ResponseModelT]]
):
    """Returns the response of calling the API of the provided decorator.

    Example:

    ```python
    import asyncio

    from mirascope.core import BasePrompt, openai, prompt_template


    @prompt_template("Recommend a {genre} book")
    class BookRecommendationPrompt(BasePrompt):
        genre: str


    async def run():
        prompt = BookRecommendationPrompt(genre="fantasy")
        response = await prompt.run_async(openai.call("gpt-4o-mini"))
        print(response.content)


    asyncio.run(run())
    ```
    """
    kwargs = {field: getattr(self, field) for field in self.model_fields}
    args_str = ", ".join(kwargs.keys())
    namespace, fn_name = {}, self.__class__.__name__
    exec(f"async def {fn_name}({args_str}): ...", namespace)
    return reduce(
        lambda res, f: f(res),  # pyright: ignore [reportArgumentType, reportCallIssue]
        [
            metadata(get_metadata(self, self.dynamic_config())),
            prompt_template(get_prompt_template(self)),
            call_decorator,
            *additional_decorators,
        ],
        namespace[fn_name],
    )(**kwargs)

prompt_template

prompt_template(
    template: str | None = None,
) -> PromptDecorator | MessagesDecorator

A decorator for setting the prompt_template of a BasePrompt or call.

Usage Documentation

Prompts

Example:

from mirascope.core import openai, prompt_template


@prompt_template()
def recommend_book(genre: str) -> str:
    return f"Recommend a {genre} book"


print(recommend_book("fantasy"))
# Output: [BaseMessageParam(role='user', content='Recommend a fantasy book')]

Returns:

Name	Type	Description
decorator	Callable	The decorator function that turns the decorated function into a prompt template.

Source code in mirascope/core/base/prompt.py

def prompt_template(
    template: str | None = None,
) -> PromptDecorator | MessagesDecorator:
    """A decorator for setting the `prompt_template` of a `BasePrompt` or `call`.

    usage docs: learn/prompts.md#prompt-templates-messages

    Example:

    ```python
    from mirascope.core import openai, prompt_template


    @prompt_template()
    def recommend_book(genre: str) -> str:
        return f"Recommend a {genre} book"


    print(recommend_book("fantasy"))
    # Output: [BaseMessageParam(role='user', content='Recommend a fantasy book')]
    ```

    Returns:
        decorator (Callable): The decorator function that turns the decorated function
            into a prompt template.
    """

    if template is None:
        # For @prompt_template() case
        decorator = messages_decorator()
        decorator.__mirascope_prompt_template__ = True  # pyright: ignore [reportAttributeAccessIssue]
        return decorator

    @overload
    def inner(
        prompt: Callable[_P, BaseDynamicConfig],
    ) -> Callable[_P, list[BaseMessageParam]]: ...

    @overload
    def inner(
        prompt: Callable[_P, Awaitable[BaseDynamicConfig]],
    ) -> Callable[_P, Awaitable[list[BaseMessageParam]]]: ...

    @overload
    def inner(prompt: type[_BasePromptT]) -> type[_BasePromptT]: ...

    def inner(
        prompt: type[_BasePromptT]
        | Callable[_P, BaseDynamicConfig]
        | Callable[_P, Awaitable[BaseDynamicConfig]],
    ) -> (
        Callable[_P, list[BaseMessageParam]]
        | Callable[_P, Awaitable[list[BaseMessageParam]]]
        | type[_BasePromptT]
    ):
        """Updates the `prompt_template` class attribute to the given value."""
        prompt._prompt_template = template  # pyright: ignore [reportAttributeAccessIssue,reportFunctionMemberAccess]

        if isinstance(prompt, type):
            return prompt

        if fn_is_async(prompt):

            @wraps(prompt)
            async def get_base_message_params_async(
                *args: _P.args, **kwargs: _P.kwargs
            ) -> list[BaseMessageParam]:
                return parse_prompt_messages(
                    roles=SUPPORTED_MESSAGE_ROLES,
                    template=template,
                    attrs=get_fn_args(prompt, args, kwargs),
                    dynamic_config=await prompt(*args, **kwargs),
                )

            get_base_message_params_async._original_fn = prompt  # pyright: ignore [reportAttributeAccessIssue,reportFunctionMemberAccess]
            return get_base_message_params_async

        else:

            @wraps(prompt)
            def get_base_message_params(
                *args: _P.args, **kwargs: _P.kwargs
            ) -> list[BaseMessageParam]:
                return parse_prompt_messages(
                    roles=SUPPORTED_MESSAGE_ROLES,
                    template=template,
                    attrs=get_fn_args(prompt, args, kwargs),
                    dynamic_config=prompt(*args, **kwargs),
                )

            get_base_message_params._original_fn = prompt  # pyright: ignore [reportAttributeAccessIssue,reportFunctionMemberAccess]
            return get_base_message_params

    inner.__mirascope_prompt_template__ = True  # pyright: ignore [reportFunctionMemberAccess]
    return inner

metadata

metadata(metadata: Metadata) -> MetadataDecorator

A decorator for adding metadata to a BasePrompt or call.

Adding this decorator to a BasePrompt or call updates the metadata annotation to the given value. This is useful for adding metadata to a BasePrompt or call that can be used for logging or filtering.

Example:

from mirascope.core import metadata, openai, prompt_template


@openai.call("gpt-4o-mini")
@prompt_template("Recommend a {genre} book")
@metadata({"tags": {"version:0001", "books"}})
def recommend_book(genre: str):
    ...


response = recommend_book("fantasy")
print(response.metadata)

Returns:

Name	Type	Description
decorator	Callable	The decorator function that updates the _metadata attribute of the decorated input prompt or call.

Source code in mirascope/core/base/prompt.py

def metadata(metadata: Metadata) -> MetadataDecorator:
    """A decorator for adding metadata to a `BasePrompt` or `call`.

    Adding this decorator to a `BasePrompt` or `call` updates the `metadata` annotation
    to the given value. This is useful for adding metadata to a `BasePrompt` or `call`
    that can be used for logging or filtering.

    Example:

    ```python
    from mirascope.core import metadata, openai, prompt_template


    @openai.call("gpt-4o-mini")
    @prompt_template("Recommend a {genre} book")
    @metadata({"tags": {"version:0001", "books"}})
    def recommend_book(genre: str):
        ...


    response = recommend_book("fantasy")
    print(response.metadata)
    ```

    Returns:
        decorator (Callable): The decorator function that updates the `_metadata`
            attribute of the decorated input prompt or call.
    """

    @overload
    def inner(prompt: type[_BasePromptT]) -> type[_BasePromptT]: ...

    @overload
    def inner(prompt: Callable[_P, _R]) -> Callable[_P, _R]: ...

    def inner(
        prompt: type[_BasePromptT] | Callable[_P, _R],
    ) -> type[_BasePromptT] | Callable[_P, _R]:
        """Updates the `metadata` class attribute to the given value."""
        prompt._metadata = metadata  # pyright: ignore [reportAttributeAccessIssue,reportFunctionMemberAccess]
        return prompt

    return inner