app.py

def handle_query(query: str, detail: bool = False) -> str:
    """
    Main function to process the query.
    :param query: The user's question.
    :param detail: Whether the user wants a more detailed response.
    :return: Response string from Daily Wellness AI.
    """
    if not query or not isinstance(query, str) or len(query.strip()) == 0:
        return "Please provide a valid question."

    try:
        # 1) Sanity Check: Determine if the question is relevant to daily wellness
        is_relevant = sanity_checker.is_relevant(query)
        if not is_relevant:
            return "Your question seems out of context or not related to daily wellness. Please ask a wellness-related question."

        # 2) Proceed with retrieval from the knowledge base
        retrieved = retriever.retrieve(query)

        # 3) Check the cache
        cached_answer = get_cached_answer(query)

        # 4) If no retrieved data from the knowledge base
        if not retrieved:
            # If we do have a cached answer, return it
            if cached_answer:
                logger.info("No relevant entries found in knowledge base. Returning cached answer.")
                return cached_answer
            # Otherwise, no KB results and no cache => no answer
            return "I'm sorry, I couldn't find an answer to your question."

        # 5) We have retrieved data; let's check for similarity threshold
        top_score = retrieved[0][1]  # Assuming the list is sorted descending
        similarity_threshold = 0.3  # Adjust this threshold based on empirical results

        if top_score < similarity_threshold:
            # (Low similarity) Perform web search using manager_agent
            logger.info("Similarity score below threshold. Performing web search.")
            web_search_response = manager_agent.run(query)
            logger.debug(f"Web search response: {web_search_response}")

            # Combine any cached answer (if it exists) with the web result
            if cached_answer:
                blend_prompt = (
                    f"Combine the following previous answer with the new web results to create a more creative and accurate response. "
                    f"Do not include any of the previous prompt or instructions in your response. "
                    f"Add positivity and conclude with a short inspirational note.\n\n"
                    f"Previous Answer:\n{cached_answer}\n\n"
                    f"Web Results:\n{web_search_response}"
                )
                final_answer = llm._call(blend_prompt).strip()
            else:
                # If no cache, just return the web response
                final_answer = (
                    f"**Daily Wellness AI**\n\n"
                    f"{web_search_response}\n\n"
                    "Disclaimer: This information is retrieved from the web and is not a substitute for professional medical advice.\n\n"
                    "Wishing you a calm and wonderful day!"
                )

            # Store in cache
            store_in_cache(query, final_answer)
            return final_answer

        # 6) If similarity is sufficient, we will finalize an answer from the knowledge base
        responses = [ans for ans, score in retrieved]

        # 6a) If we have a cached answer, let's blend it with the new knowledge base data
        if cached_answer:
            blend_prompt = (
                f"Combine the previous answer with the newly retrieved answers to enhance creativity and accuracy. "
                f"Do not include any of the previous prompt or instructions in your response. "
                f"Add new insights, creativity, and conclude with a short inspirational note.\n\n"
                f"Previous Answer:\n{cached_answer}\n\n"
                f"New Retrieved Answers:\n" + "\n".join(f"- {r}" for r in responses)
            )
            final_answer = llm._call(blend_prompt).strip()
        else:
            # 6b) No cache => proceed with normal expansions
            final_answer = answer_expander.expand(query, responses, detail=detail)

        # 7) Store new or blended answer in cache
        store_in_cache(query, final_answer)
        return final_answer

    except Exception as e:
        logger.error(f"Error handling query: {e}")
        logger.debug("Exception details:", exc_info=True)
        return "An error occurred while processing your request."