emergingtrajectories.utils
==========================

.. py:module:: emergingtrajectories.utils


Attributes
----------

.. autoapisummary::

   emergingtrajectories.utils._extract_prediction_prompt
   emergingtrajectories.utils._extract_prediction_prompt_error


Classes
-------

.. autoapisummary::

   emergingtrajectories.utils.UtilityHelper


Functions
---------

.. autoapisummary::

   emergingtrajectories.utils.is_numeric
   emergingtrajectories.utils.run_forecast


Module Contents
---------------

.. py:data:: _extract_prediction_prompt
   :value: Multiline-String

   .. raw:: html

      <details><summary>Show Value</summary>

   .. code-block:: python

      """You are helping a researcher with a data extraction exercise. You will be provided with a prediction statement and a broader piece of text. Your objective is to extract the specific numerical prediction and provide it as a response. DO NOT qualify your response in any way.
      
      For example, suppose you have the following...
      
      |||---|||
      PREDICTION STATEMENT: please predict the probability that the price of Bitcoin will exceed $100,000 by the end of 2024.
      
      TEXT: The probability that bitcoin will exceed $100,000 by the end of 2024 is 0.37.
      |||---|||
      
      In the case above, your response would simply be "0.37".
      
      The actual metrics (i.e., prediction) might be provided with formatting. For example...
      
      |||---|||
      PREDICTION STATEMENT: The probability that Boeing's (NYSE:BA) share price at the close of markets on or before March 1, 2024 will be $220.00 USD or higher is _____ (value between 0.00 and 1.00).
      
      TEXT: The probability that Boeing's (NYSE:BA) share price at the close of markets on or before March 1, 2024, will be $220.00 USD or higher is **0.65**.
      |||---|||
      
      In this case, ignore the asterisks or instructions ("value between 0.00 and 1.00") and provide the correct response, which is 0.65.
      
      The user will provide you with a PREDICTION STATEMENT and TEXT and you need to answer like the above.
      
      On the extremely rare occasion that the TEXT does not have a proper numerical prediction or you are unable to extract it, simply respond with "UNCLEAR".
      """

   .. raw:: html

      </details>



.. py:data:: _extract_prediction_prompt_error
   :value: 'UNCLEAR'


.. py:function:: is_numeric(string: str) -> bool

   Checks whether the 'string' passed as an argument can be converted into a numeric value.

   :param string: the string in question

   :returns: Boolean value; True if the string can be converted into a numeric value, False otherwise.


.. py:function:: run_forecast(function_to_call, n, *args, **kwargs)

.. py:class:: UtilityHelper(api_key, model='gpt-4-0125-preview')

   Bases: :py:obj:`object`


   The UtilityHelper class is used to extract predictions from text messages.

   :param api_key: the OpenAI API key
   :param model: the OpenAI model to use for the extraction process


   .. py:attribute:: api_key


   .. py:attribute:: model
      :value: 'gpt-4-0125-preview'



   .. py:method:: extract_prediction(response: str, statement_challenge: str) -> float

      Extracts the prediction value from the response to a statement challenge.

      :param response: the response to the statement challenge (i.e., what was predicted by another LLM)
      :param statement_challenge: the statement challenge -- what is being predicted

      :returns: The extracted prediction value as a float. Raises an exception if the prediction cannot be extracted.