Untitled

def llm_generate_insights(remake_df, model, generation_config, safety_config):
    try:
        structured_data = remake_df.to_dict(orient='records')

        # Step 1: Generate trends and patterns
        trends_prompt = f"""
        Analyze the following data of RPG remakes/remasters:
        {structured_data}
        Provide insights on the most common features, patterns, and player reception trends.
        Summarize in a structured format.
        """

        logging.info("First Call: Generating trends and patterns...")
        trends_response = model.generate_content(
            contents=trends_prompt, 
            generation_config=generation_config, 
            safety_settings=safety_config
        )
        time.sleep(2)

        # Step 2: Generate Chrono Trigger recommendations
        chrono_trigger_prompt = f"""
        Based on the following trends and RPG remakes/remaster data, suggest the top 5 most relevant RPG remakes/remasters for a Chrono Trigger remake. Explain why these titles are relevant.
        Also, suggest actionable features or patterns that should be included in the remake based on successful games.

        Trends and Patterns:
        {trends_response.text}

        RPG Data:
        {structured_data}
        """

        logging.info("Second Call: Generating Chrono Trigger recommendations...")
        chrono_response = model.generate_content(
            contents=chrono_trigger_prompt, 
            generation_config=generation_config, 
            safety_settings=safety_config
        )
        time.sleep(2)

        # Step 3: Extract feature patterns for visualization
        features_prompt = f"""
        Extract the most mentioned features or patterns from the RPG remakes/remasters dataset. 
        Provide a concise list of features with frequency counts in JSON format - PICK MAX 20.
        {structured_data}
        
        Sample Format:
        {"features": {"Improved visuals": 2, "New game modes": 2, "Gameplay enhancements": 25, "Expanded narrative": 24}}
        """

        logging.info("Third Call: Extracting features for visualization...")
        features_response = model.generate_content(
            contents=features_prompt, 
            generation_config=generation_config, 
            safety_settings=safety_config
        )
        time.sleep(2)

        # Step 4: Extract relevant titles
        titles_prompt = f"""
        Based on the following response for Chrono Trigger recommendations:
        {chrono_response.text}

        Provide the top 5 titles in JSON format with the following fields:
        - Title
        - Features: Relevant to Chrono Trigger Remake (up to 3 features)
        """

        logging.info("Fourth Call: Extracting relevant titles...")
        titles_response = model.generate_content(
            contents=titles_prompt, 
            generation_config=generation_config, 
            safety_settings=safety_config
        )
        time.sleep(2)

        # Parse LLM responses using parse_llm_response
        trends_summary = trends_response.text.strip()
        chrono_trigger_summary = chrono_response.text.strip()
        feature_insights = parse_llm_response(features_response.text)
        relevant_titles = parse_llm_response(titles_response.text)

        if feature_insights is None or relevant_titles is None:
            logging.error("Failed to parse some of the LLM responses.")
            return {}

        logging.info("All LLM calls and parsing completed successfully.")
        return {
            "trends_summary": trends_summary,
            "chrono_trigger_summary": chrono_trigger_summary,
            "feature_insights": feature_insights,
            "relevant_titles": relevant_titles
        }

    except Exception as e:
        logging.error(f"Error in llm_generate_insights: {e}")
        return {}