util_service.py

from langchain_core.prompts import ChatPromptTemplate
import numpy as np

keywords = ["i encountered", "429", "unfortunately", "unexpected character", "unsupported", "error", "sorry", "response", "unable", "because", "too many"]

def contains_keywords(text, keywords):
    return any(keyword.lower() in text.lower() for keyword in keywords)

def process_answer(answer):
    # Handle empty list scenario
    if isinstance(answer, list) and not answer:
        return "Empty response received."

    # Handle dictionary with possible "answer" or "error"
    if isinstance(answer, dict):
        if "answer" in answer and contains_keywords(answer["answer"], keywords):
            return True
        if "error" in answer and contains_keywords(answer["error"], keywords):
            return True

    # Default case if no keywords are found
    return False


def _prompt_generator(question: str, chart_required: bool, csv_url: str):

    chat_prompt = f"""You are a senior data analyst working with CSV data. Adhere strictly to the following guidelines:  

                         1. **Data Verification:** Always inspect the data with `.sample(5).to_dict()` before performing any analysis.  
                         2. **Data Integrity:** Ensure proper handling of null values to maintain accuracy and reliability.  
                         3. **Communication:** Provide concise, professional, and well-structured responses.
                         4. Avoid including any internal processing details or references to the methods used to generate your response (ex: based on the tool call, using the function -> These types of phrases.)
                         5. Always use pd.read_csv({csv_url}) to read the CSV file.
                         6. **Full Dataset Interpretation Rule:** When asked about the entire dataset, provide a concise summary of the dataset's structure, such as the number of rows, columns, and data types.
                            (Example: “give me total number of columns”, “show all statistics”, “full summary”, etc.)

                         **Query:** {question}  
         
                    """               
        
    chart_prompt = f"""You are a senior data analyst working with CSV data. Follow these rules STRICTLY:

                       1. Generate ONE unique identifier FIRST using: unique_id = uuid.uuid4().hex
                       2. Visualization requirements:
                       - Adjust font sizes, rotate labels (45° if needed), truncate for readability
                       - Figure size: (12, 6)
                       - Descriptive titles (fontsize=14)
                       - Colorblind-friendly palettes
                       - Do not use any visualization library other than matplotlib or seaborn
                       3. File handling rules:
                       - Create MAXIMUM 2 charts if absolutely necessary
                       - For multiple charts:
                           * Arrange in grid format (2x1 vertical layout preferred)
                           * Use SAME unique_id with suffixes: 
                           - f"{{unique_id}}_1.png"
                           - f"{{unique_id}}_2.png"
                       - Save EXCLUSIVELY to "generated_charts" folder
                       - File naming: f"chart_{{unique_id}}.png" (for single chart)
                       4. FINAL OUTPUT MUST BE:
                       - For single chart: f"generated_charts/chart_{{unique_id}}.png"
                       - For multiple charts: f"generated_charts/chart_{{unique_id}}.png" (combined grid image)
                       - ONLY return this full path string, nothing else

                       **Query:** {question}

                       IMPORTANT: 
                       - Generate the unique_id FIRST before any operations
                       - Use THE SAME unique_id throughout entire process
                       - NEVER generate new UUIDs after initial creation
                       - Return EXACT filepath string of the final saved chart
                       - Always use pd.read_csv({csv_url}) to read the CSV file
                       """

        
    if chart_required:
        return ChatPromptTemplate.from_template(chart_prompt)
    else:
        return ChatPromptTemplate.from_template(chat_prompt)
    
    
    
def handle_out_of_range_float(value):
    if isinstance(value, float):
        if np.isnan(value):
            return None
        elif np.isinf(value):
            return "Infinity"
    return value