promptu

Promptu

Write dynamic LLM prompts as natural language

Promptu provides a suite of natural language generation utilities for prompt engineers, such as truncation, pluralization and natural language list formatting.

Project Status

️ This project is currently in its infancy. Please report bugs and expect breaking changes.

Installation

pip install promptu

Example

from promptu import join

def find_matching_color(existing_colors):
    return f'What color goes well with {join(items=existing_colors, conjunction="and")}?'

# Prints "What color goes well with blue, purple and white?"
print(find_matching_color(['blue', 'purple', 'white']))

Using in Langchain

runnable = (
    {'prompt': RunnableLambda(find_matching_color)}
    | PromptTemplate.from_template('{prompt}')
    | model
)
runnable.invoke(['blue', 'purple', 'white'])

Prompting Utilities

join()

Formats a list of items as a natural language list.

Syntax:

join(items: Sequence, conjunction: str) -> str

pluralize()

Selects the singular or plural form of a word based on the number of items.

Syntax:

pluralize(singular: str, plural: str, items: Sized) -> str

truncate()

Truncates a string to a maximum length. If the text is longer than the maximum length, all characters after the maximum length are replaced with the suffix.

Syntax:

truncate(
    text: str, max_length: int, mode=TruncateMode.CHARACTER, suffix="..."
) -> str

Contributing

PRs are welcome! Please add or update the relevant unit tests and format the project with black .