transferring files from spendtracker

LLM_openai.py

@@ -0,0 +1,31 @@
+from openai import OpenAI
+import os
+my_key = os.environ.get('MY_OPENAI_KEY')
+client = OpenAI(
+    api_key= my_key
+)
+
+def expense_classifier(expense):
+    chat_completion = client.chat.completions.create(
+    messages=[
+        {   "role": "system",
+            "content": [
+        {
+          "type": "text",
+          "text": """You are a helpful personal finance assistant. 
+          The user will input some expense concept and you will classify it in a broader category from the following:
+            [alcohol, food, restaurant, clothing, entertainment, transport, sports, wellbeing, personal_development,others] 
+            Provide only the category as an answer
+            """
+        }
+      ]
+        },
+        {
+            "role": "user",
+            "content": expense,
+        }
+    ],
+    model="gpt-4o-mini",
+    temperature=0
+)
+    return chat_completion.choices[0].message.content

app.py

@@ -0,0 +1,102 @@
+import re
+import os
+import gspread
+import gradio as gr
+import datetime
+import pandas as pd
+import matplotlib
+matplotlib.use('Agg')
+import matplotlib.pyplot as plt
+from LLM_openai import client, expense_classifier
+from utils import create_plot, create_barplot
+from dataframe_processing import dataframe_process
+from vision_api_call import process_image
+
+#connect to the service account
+gc = gspread.service_account(filename="spend_tracker_creds.json")
+#connect to your sheet (between "" = the name of your G Sheet, keep it short)
+spreadsheet = gc.open("spend_tracker").sheet1
+
+def update_spend_from_image(img):
+    ##This processes the image
+    extracted_dictionay_from_image=process_image(img)
+    total_from_receipt=extracted_dictionay_from_image['total']
+    concept_from_receipt=extracted_dictionay_from_image['purchase_summary']
+    shop_type=extracted_dictionay_from_image['store_type']
+    shop_name=extracted_dictionay_from_image['store']
+    receipt_items=str(extracted_dictionay_from_image['items'])
+
+    concept_and_shop=concept_from_receipt+" from "+shop_type+f" ({shop_name})"
+    
+    ##The function update_spend in the line below only takes a string and float
+    category, day_month, todays_amount, current_week_amount , fig, fig2, fig3=update_spend(concept_and_shop, total_from_receipt, receipt_items)
+
+    return category, day_month, todays_amount, current_week_amount , fig, fig2, fig3
+
+
+
+def update_spend(concept, amount, items_from_receipt=None):
+    category=expense_classifier(concept)
+    today = datetime.date.today()
+    
+    # Append a new row
+    spreadsheet.append_row([str(today), concept,float(amount),category, items_from_receipt ])
+    
+    day_month, todays_amount, current_week_amount , fig, fig2, fig3=dataframe_process(spreadsheet)
+    return category, day_month, todays_amount, current_week_amount , fig, fig2, fig3
+
+def show_plots():
+    category='N/A'
+    today = 'N/A'
+    
+    # Append a new row
+    # spreadsheet.append_row([str(today), concept,float(amount),category, items_from_receipt ])
+    
+    day_month, todays_amount, current_week_amount , fig, fig2, fig3=dataframe_process(spreadsheet)
+    return category, day_month, todays_amount, current_week_amount , fig, fig2, fig3
+
+
+
+
+with gr.Blocks() as demo:
+    gr.Markdown("Expense tracker")
+    
+    with gr.Tab("Manual input"):
+        with gr.Row():
+            concept = gr.Textbox(label="concept")
+        with gr.Row():
+            amount= gr.Textbox(label="amount")
+        btn_manual = gr.Button("Submit expense")
+
+    with gr.Tab("Upload receipt"):
+        with gr.Row():
+            input_image = gr.Image( type="pil")
+        btn_image = gr.Button("Submit expense")
+        
+        
+        
+    
+    
+    with gr.Row():
+        btn_show = gr.Button("Show plots")
+
+    with gr.Row():
+        expense_class= gr.Textbox(label='expense class')
+        ui_date=gr.Textbox(label='date')
+        expenses_today= gr.Textbox(label='expenses today')
+        expenses_this_week= gr.Textbox(label='expenses this week')
+    with gr.Row():
+        daily_plot=gr.Plot(label=None)
+        week_plot=gr.Plot()
+        category_plot=gr.Plot()
+
+
+    btn_manual.click(fn=update_spend, inputs=[concept, amount], outputs=[expense_class,ui_date, expenses_today,expenses_this_week,
+                                                                  daily_plot,week_plot,category_plot])
+    btn_image.click(fn=update_spend_from_image, inputs=[input_image], outputs=[expense_class,ui_date, expenses_today,expenses_this_week,
+                                                                  daily_plot,week_plot,category_plot])
+    btn_show.click(fn=show_plots, inputs=[], outputs=[expense_class,ui_date, expenses_today,expenses_this_week,
+                                                                  daily_plot,week_plot,category_plot])
+
+
+demo.launch()

dataframe_processing.py

@@ -0,0 +1,42 @@
+import pandas as pd
+import re
+from utils import create_plot, create_barplot
+
+def dataframe_process(sheet):
+    # Create dataframe
+    df = pd.DataFrame(sheet.get_all_records())
+    df['date'] = pd.to_datetime(df['date'])
+    df['week'] = df['date'].dt.to_period('W')
+
+    #Per day
+    daily_total=df[['concept','amount','category','date']].groupby('date').sum().reset_index()
+    daily_total['month_day'] = daily_total['date'].dt.strftime('%m-%d')
+
+    
+
+    fig = create_plot(daily_total['month_day'],daily_total['amount'])
+
+    daily_total=daily_total.sort_values(by='date', ascending=False)
+    daily_total=daily_total[['date', 'amount']]
+    todays_amount=daily_total.iloc[0]['amount']
+    todays_date=daily_total.iloc[0]['date']
+    day_month = todays_date.strftime('%m-%d-%Y')
+
+    #Per week
+    # Group by 'week' and sum the values
+    weekly_df = df[['concept','amount','category','week']].groupby('week').sum().reset_index()
+    weekly_df=weekly_df.sort_values(by='week', ascending=True)
+    # weekly_df.index=weekly_df.index.strftime('%m-%d')
+    current_week_amount="{:.2f}".format(weekly_df.iloc[-1]['amount'])
+    weekly_df['week'] = weekly_df['week'].astype(str)
+    weekly_df['week']=weekly_df['week'].apply(lambda x: re.sub(r'\d{4}-', '', x))
+
+    fig2 = create_barplot(weekly_df['week'],weekly_df['amount'],"week","amount","expense per week")
+
+
+    #Per category
+    expenses_per_category=df[['concept','amount','category']].groupby('category').sum().reset_index()
+    expenses_per_category=expenses_per_category.sort_values(by='amount', ascending=False)
+    fig3 = create_barplot(expenses_per_category['category'],expenses_per_category['amount'],"category","amount","expense per category")
+
+    return day_month, todays_amount, current_week_amount , fig, fig2, fig3

utils.py

@@ -0,0 +1,70 @@
+import matplotlib.pyplot as plt
+from LLM_openai import client, expense_classifier
+import datetime
+import base64
+from io import BytesIO
+import re
+import json
+import matplotlib.dates as mdates
+
+def create_plot(x, y):
+    fig, ax = plt.subplots()
+    ax.plot(x, y, marker='o')
+    for i in range(len(x)):
+        ax.text(x[i], y[i], f'{y[i]:.2f}', fontsize=10, ha='left', va='bottom')
+    ax.set_xlabel('Money')
+    ax.set_ylabel('Expenses')
+    ax.set_title('Daily expenses')
+    ax.set_xticks(x[::5])  # Show every 5th day
+    ax.set_xticklabels(x[::5], rotation=45, ha='right') 
+    plt.xticks(rotation=30)
+    return fig
+
+
+def create_barplot(x, y, xlabel,ylabel, title):
+    fig, ax = plt.subplots()
+    ax.bar(x, y)
+    for i in range(len(x)):
+        ax.text(x[i], y[i], f'{y[i]:.2f}', fontsize=10, ha='left', va='bottom')
+    ax.set_xlabel(xlabel)
+    ax.set_ylabel(ylabel)
+    ax.set_title(title)
+    plt.xticks(rotation=30)
+    return fig
+
+
+####
+##OCR functions
+
+
+
+def pil_to_base64(pil_img):
+    img_buffer = BytesIO()
+    pil_img.save(img_buffer, format='JPEG')
+    byte_data = img_buffer.getvalue()
+    base64_str = base64.b64encode(byte_data).decode("utf-8")
+    return base64_str
+
+def js_to_prefere_the_back_camera_of_mobilephones():
+    custom_html = """
+    <script>
+    const originalGetUserMedia = navigator.mediaDevices.getUserMedia.bind(navigator.mediaDevices);
+    
+    navigator.mediaDevices.getUserMedia = (constraints) => {
+      if (!constraints.video.facingMode) {
+        constraints.video.facingMode = {ideal: "environment"};
+      }
+      return originalGetUserMedia(constraints);
+    };
+    </script>
+    """
+    return custom_html
+
+def result_cleaner(text):
+    pattern = r'\{[^}]*\}'
+    match = re.search(pattern, text)
+    match_string=match[0]
+    json_dict=json.loads(match_string)
+
+
+    return json_dict

vision_api_call.py

@@ -0,0 +1,61 @@
+from openai import OpenAI
+import os
+import base64
+from io import BytesIO
+from PIL import Image
+from utils import pil_to_base64, result_cleaner
+
+
+my_key = os.environ.get('MY_OPENAI_KEY')
+client = OpenAI(
+    api_key= my_key
+)
+
+prompt="""
+
+You'll be analyzing purchase receipts you will extract the following information:
+-date: date of the purchase in the format YYYY-MM-DD
+-Store: Name of the store where items or services were purchased
+-Store_type: Type of store (supermarket, restaurant, bookstore, etc)
+-Purchase_summary: in maximum 5 words summarize the purchase (cleaning products, breakfast, books, food items for home, online services, clothes ,etc)
+-items: create a list of all the items in the receipt
+-total: total amount spent
+
+Provide your answer in a dictionary like the following
+{{date: "xxxx-xx-xx"
+store: "example store",
+store_type: "supermarket",
+purchase_summary: "food items for home",
+items: "xxxx, xxxx, xxx"
+total: xxxx }}
+"""
+
+
+def analyse_image(processed_image):
+    #
+    response = client.chat.completions.create(
+        model="gpt-4o-mini",
+        messages=[
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "text",
+                        "text": prompt,
+                    },
+                    {
+                        "type": "image_url",
+                        "image_url": {"url": f"data:image/jpeg;base64,{processed_image}"},
+                    },
+                ],
+            }
+        ],
+    )
+    return response.choices[0].message.content
+
+def process_image(img):
+    image_base64=pil_to_base64(img)
+    analysis_result=analyse_image(image_base64)
+    clean_analysis_result=result_cleaner(analysis_result)
+ 
+    return clean_analysis_result