def categorize_transactions(transaction_names, llm):

    prompt = f"""把以下费用分到适当的类别中。

请记住，类别应从以下列表中选择一个，根据它们的主要目的或性质选择最相关的类别：{categories_string}。\n

输出格式应始终为：transaction name - category。例如：Spotify #2 - Entertainment, Basic Fit Amsterdam Nld #3 - Fitness/Sports \n

以下是待分类的交易：{transaction_names} \n"""

    print(prompt)

    filtered_response = []

    # retry is the LLM output is not consistent

    while len(filtered_response) < 2:

        response = llm.invoke(prompt).split("\n")

        print(response)

        # Remove items that do not contain "transaction: category" pairs

        filtered_response = [item for item in response if '-' in item]

    print(filtered_response)

    # Put in dataframe

    categories_df = pd.DataFrame({"Transaction vs category": filtered_response})

    size_dif = len(categories_df) - len(transaction_names.split(","))

    if size_dif >= 0:

        categories_df["Transaction"] = transaction_names.split(",") + [None] * size_dif

    else:

        categories_df["Transaction"] = transaction_names.split(",")[:len(categories_df)]

    categories_df["Category"] = categories_df["Transaction vs category"].str.split("-", expand=True)[1]

    return categories_df

def hop(start, stop, step):

    for i in range(start, stop, step):

        yield i

    yield stop

def process_data(df: pd.DataFrame):

    unique_transactions = df["Name/Description"].unique()

    index_list = list(hop(0, len(unique_transactions), 30))

    # Intialise the categories_df_all dataframe

    categories_df_all = pd.DataFrame()

    # Loop through the index_list

    for i in range(0, len(index_list) - 1):

        print(f"Looping: {i}")

        transaction_names = unique_transactions[index_list[i] : index_list[i + 1]]

        transaction_names = ",".join(transaction_names)

        categories_df = categorize_transactions(transaction_names, llm)

        categories_df_all = pd.concat(

            [categories_df_all, categories_df], ignore_index=True

        )

    # futher clean data:

    # Drop NA values

    categories_df_all = categories_df_all.dropna()

    # Remove the numbering eg "1. " from Transaction column

    categories_df_all["Transaction"] = categories_df_all["Transaction"].str.replace(

        r"\d+\.\s?", "", regex=True

    ).str.strip()

    new_df = pd.merge(

        df,

        categories_df_all,

        left_on="Name/Description",

        right_on="Transaction",

        how="left",

    )

    new_df.to_csv(f"data/{uploaded_file.name}_categorized.csv", index=False)

    return new_df

def financial_analysis(data:pd.DataFrame):

    key_figures = {}

    # Calculate yearly total income and total expenses

    yearly_income = data.loc[data['Expense/Income'] == 'Income'].groupby('Year')['Amount(EUR)'].sum().mean()

    yearly_expenses = data.loc[data['Expense/Income'] == 'Expense'].groupby('Year')['Amount(EUR)'].sum().mean()

    # Identify the top expense categories

    top_expenses = data.loc[data['Expense/Income'] == 'Expense'].groupby('Category')['Amount(EUR)'].sum().sort_values(

        ascending=False)

    # Calculate average monthly income and expenses

    monthly_income = data.loc[data['Expense/Income'] == 'Income'].groupby(data['Date'].dt.to_period('M'))[

        'Amount(EUR)'].sum().mean()

    monthly_expenses = data.loc[data['Expense/Income'] == 'Expense'].groupby(data['Date'].dt.to_period('M'))[

        'Amount(EUR)'].sum().mean()

    # Determine the savings rate

    savings = yearly_income - yearly_expenses

    savings_rate = (savings / yearly_income) * 100 if yearly_income > 0 else 0

    key_figures['Average Annual Income'] = f"€{yearly_income:,.2f}"

    key_figures['Average Annual Expenses'] = f"€{yearly_expenses:,.2f}"

    key_figures['Annual Savings Rate'] = f" {savings_rate:.2f}%"

    key_figures['Top Expense Categories'] = {category: f"€{amount:,.2f}" for category, amount in

                                             top_expenses.head().items()}

    key_figures['Average Monthly Income'] = f"€{monthly_income:,.2f}"

    key_figures['Average Monthly Expenses'] = f"€{monthly_expenses:,.2f}"

    return key_figures

def plot_income_vs_expense_over_time(df):

    # Income vs Expense Over time

    st.markdown("1. Income vs Expense Over time")

    income_expense_summary = (

        df.groupby(["YearMonth", "Expense/Income"])["Amount(EUR)"]

        .sum()

        .unstack()

        .fillna(0)

    )

    income_expense_summary.plot(kind="bar", figsize=(10, 8))

    plt.title("Income vs Expenses Over Time")

    plt.ylabel("Amount (EUR)")

    plt.xlabel("Month")

    plt.savefig("data/income_vs_expense_over_time.png", bbox_inches="tight")

    st.pyplot(plt)

def plot_saving_rate_trend(data: pd.DataFrame):

    st.markdown("2. Monthly Saving Rate Trend")

    monthly_data = data.groupby(['YearMonth', 'Expense/Income'])['Amount(EUR)'].sum().unstack().fillna(0)

    monthly_data['Savings Rate'] = (monthly_data['Income'] - monthly_data['Expense']) / monthly_data['Income'] * 100

    fig, ax = plt.subplots()

    monthly_data['Savings Rate'].plot(ax=ax)

    ax.set_xlabel('Month')

    ax.set_ylabel('Savings Rate (%)')

    plt.savefig("data/saving_rate_over_time.png", bbox_inches="tight")

    st.pyplot(fig)

def plot_income_source_analysis(data: pd.DataFrame):

    st.markdown("3. Income Sources Analysis")

    income_sources = data[data['Expense/Income'] == 'Income'].groupby('Category')['Amount(EUR)'].sum()

    income_sources.plot(kind="pie", figsize=(10, 8), autopct="%1.1f%%", startangle=140)

    plt.title("Income Sources Analysis")

    plt.ylabel("")  # Hide the y-label as it's unnecessary for pie charts

    plt.savefig("data/income_source_analysis.png", bbox_inches="tight")

    st.pyplot(plt)

def plot_category_wise_spending_analysis(data: pd.DataFrame):

    st.markdown("4. Category-wise Spending Analysis")

    expenses_by_category = data[data['Expense/Income'] == 'Expense'].groupby('Category')['Amount(EUR)'].sum()

    expenses_by_category.plot(kind="pie", figsize=(10, 8), autopct="%1.1f%%", startangle=140)

    plt.title("Expenses Analysis")

    plt.ylabel("")  # Hide the y-label as it's unnecessary for pie charts

    plt.savefig("data/expense_category_analysis.png", bbox_inches="tight")

    st.pyplot(plt)

st.title("My Local AI Finance Insighter")

st.markdown(

    "**A personalized and secure approach to analyzing financial data, providing insights and recommendations tailored to individual needs.**"

)

analysis_results = financial_analysis(total_df)

results_str = ""

# Loop through the dictionary

for key, value in analysis_results.items():

    if isinstance(value, dict):

        # If the value is another dictionary, further iterate to get sub-keys and values

        sub_results = ', '.join([f"{sub_key}: {sub_value}" for sub_key, sub_value in value.items()])

        results_str += f"{key}: {sub_results}\n"

    else:

        # For direct key-value pairs, simply concatenate

        results_str += f"{key}: {value}\n"

st.subheader("Yearly Figures")

col1, col2, col3 = st.columns(3)

col1.metric(label="Average Annual Income", value=analysis_results['Average Annual Income'])

col2.metric(label="Average Annual Expenses", value=analysis_results['Average Annual Expenses'])

col3.metric(label="Savings Rate", value=analysis_results['Annual Savings Rate'])

# Display average monthly figures

st.subheader("Average Monthly Figures")

col1, col2 = st.columns(2)

col1.metric(label="Average Monthly Income", value=analysis_results['Average Monthly Income'])

col2.metric(label="Average Monthly Expenses", value=analysis_results['Average Monthly Expenses'])

# Display top expense categories in a table

st.subheader("Top Expense Categories")

expenses_df = pd.DataFrame(list(analysis_results['Top Expense Categories'].items()), columns=['Category', 'Amount'])

st.table(expenses_df)

with st.container():

    col1, col2 = st.columns(2)

    with col1:

        plot_income_vs_expense_over_time(total_df)

    with col2:

        plot_saving_rate_trend(total_df)

with st.container():

    col3, col4 = st.columns(2)

    with col3:

        plot_income_source_analysis(total_df)

    with col4:

        plot_category_wise_spending_analysis(total_df)

with st.container():

    col3, col4 = st.columns(2)

    with col3:

        plot_income_source_analysis(total_df)

    with col4:

        plot_category_wise_spending_analysis(total_df)

with st.spinner("Generating reports ..."):

    total_response = ""

    for root, dirs, files in os.walk("data"):

        for file in files:

            if file.endswith(".png"):

                response = llm_llava.invoke(

                    f"Act as an expert finance planner and analyse the image : {os.path.join(root, file)}. You should give your insights extracted from the image and key figures you see from the image "

                )

                total_response += response

    total_response += f"\nHere are the user key financial figures : {results_str}"

    st.write("---------------")

    st.markdown("**Finance analysis and budget planner**")

    summary = llm.invoke(

        f"You are a helpful and expert finance planner. Base on the following analysis: {total_response}, make a summary of the financial status of the user and suggest tips on savings. Highlight categories where the user can potentially reduce expenses and suggest an ideal savings rate based on their income and goals. Tailor these suggestions to fit the user’s lifestyle and financial objectives. Use a friendly tone. "

    )

    st.write(summary)

    st.write("---------------")

    st.markdown("**Investment tips**")

    if "user_answers_str" in st.session_state:

        user_investment_answer = st.session_state.user_answers_str

    else:

        user_investment_answer = ""

    investment_tips = llm.invoke(

        f"You are a helpful and expert finance planner. Based on the user's risk tolerance and investment goals, provide an overview of suitable investment options. Discuss the basics of stocks, bonds, mutual funds, ETFs, and other investment vehicles that align with their profile. Explain the importance of diversification and the role of risk management in investing. Offer to guide them through setting up a diversified investment portfolio, suggesting steps to get started based on their current financial situation. Use a friendly tone. Below are the user´s investment objective and risk tolerance : {user_investment_answer}"

    )

    st.write(investment_tips)