Final Response

The response from the tool used to process the final instruction may not be human-friendly, only containing a single number or such.

Therefore, the response must be passed through another tool to format it properly.

def final(llm, query, answer):
    template = """
    You have been given the following query and you know its answer as well. Formulate an appropriate response to the query based on this answer.

    ---
    ##Instructions
    1. Correct any spelling and grammatical (punctuation, etc) errors.
    <any other instructions>
    ---

    ---
    ##Examples
    <appropriate examples>
    ---
    ##Query
    {query}

    ##Answer
    {answer}
    ---
    """
    prompt_template = PromptTemplate(input_variables=["query", "answer"], template=template)
    return llm.invoke(prompt_template.format(query=query, answer=answer)).content