Upload 2 files

app.py

@@ -0,0 +1,196 @@
+import gradio as gr
+from sqlalchemy import create_engine
+import pandas as pd
+import openai
+import os
+from lida import Manager, TextGenerationConfig, llm
+from llmx.generators.text.openai_textgen import OpenAITextGenerator
+from langchain_openai import AzureChatOpenAI
+from langchain_core.runnables import RunnablePassthrough
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_core.output_parsers import StrOutputParser
+import pandas as pd
+import base64
+import numpy as np
+from PIL import Image
+from langchain_core.messages import HumanMessage
+from langchain_openai import ChatOpenAI
+import base64
+
+os.environ["AZURE_OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY")
+os.environ["AZURE_OPENAI_API_VERSION"] = "2023-06-01-preview"
+os.environ["AZURE_OPENAI_ENDPOINT"] = os.getenv("AZURE_OPENAI_ENDPOINT")
+db_host = os.getenv('DB_HOST')
+db_name = os.getenv('DB_NAME')
+db_user = os.getenv('DB_USER')
+db_password = os.getenv('DB_PASSWORD')
+
+model = AzureChatOpenAI(
+    deployment_name="CapSuiteGPT4omini",
+    openai_api_version=os.getenv("AZURE_OPENAI_API_VERSION"),
+)
+
+def choose_table(question):
+    try:
+        connection_string = f'postgresql+psycopg2://{db_user}:{db_password}@{db_host}/{db_name}'
+        engine = create_engine(connection_string)
+        capsuite_ref = 'foodBeverageSample1'
+
+
+        model = AzureChatOpenAI(
+            deployment_name="CapSuiteGPT4omini",
+            openai_api_version=os.getenv("AZURE_OPENAI_API_VERSION"),
+        )
+        
+        table_format = """
+                1.table name:cdp_sale_order,
+                its columns:trxn_id,member_id,staff_id,subsidiary_name,staff_name,team_name,trxn_ref,trxn_channel,trxn_date,trxn_year,trxn_month,trxn_day,trxn_week,remark.
+                
+                2.table name:cdp_sale_order_line,
+                its columns:trxn_item_id,trxn_id,trxn_item_target_curr_unit_price,
+                trxn_item_qty,trxn_item_discount_amt,trxn_original_net_currency,trxn_date,trxn_channel,staff_name,staff_id,member_id,display_name,pord_sku,prod_category,prod_type,prod_name,
+                capsuite_ref.
+                
+                3.table name:cdp_stock_quant,
+                its columns:stock_quant_id,prod_id,location_id,stock_quantity,stock_quantity_reserved,stock_quant_create_date,capsuite_ref.
+                """
+
+        prompt = ChatPromptTemplate.from_template("Base on the question:{question},"
+                                                "And the following table format:{table_format},"
+                                                "Dont write a complex query. Only select statement like 'select * from table_name'."
+                                                "Dont add any condition or filter to the query. The query should be generic and should return all the data from the table."
+                                                "Select all the columns from the table. "
+                                                "Only output one SQL Query without any other information even the '''sql''' prefix. ")
+
+        chain = (
+                {"question": RunnablePassthrough(), "table_format": RunnablePassthrough()}
+                # {"table_format": RunnablePassthrough()}
+                | prompt
+                | model
+                | StrOutputParser()
+        )
+        # query = 'select * from cdp_membership_summary;'
+        query = chain.invoke({"question": question, "table_format": table_format})
+        query = query.replace(f"`", '')
+        query = query.replace(f"sql", '')
+        query = query.split(';')[0] + f' where capsuite_ref = \'{capsuite_ref}\';'
+        df_data = pd.read_sql(query, engine)
+        print(f'*'*50)
+        print(f"Query: {query}")
+
+        if 'cdp_sale_order_line' in query:
+            df_data['sales_amount'] = df_data['trxn_item_target_curr_unit_price'].astype(float) * df_data['trxn_item_qty'].astype(float)
+            df_data.rename(columns={'trxn_item_target_curr_unit_price':'unit_price'}, inplace=True)
+            df_data.rename(columns={'display_name':'customer_name'}, inplace=True)
+            df_data = df_data[['trxn_item_id','trxn_id','sales_amount','unit_price','trxn_item_qty','trxn_item_discount_amt','trxn_date','trxn_channel','staff_name','customer_name','prod_category','prod_type','prod_name','capsuite_ref']]
+        
+    except Exception as e:
+        print(f"Error while: {e}")
+
+    finally:
+        engine.dispose()
+        return df_data
+
+
+# Function to encode the image
+def encode_image(image_path):
+    with open(image_path, "rb") as image_file:
+        return base64.b64encode(image_file.read()).decode('utf-8')
+
+
+def random_response(message, history):
+    df_data = choose_table(message)
+
+    question = message
+    # fill na with empty string
+    df_data.fillna('', inplace=True)
+    # loop columns, if column is object type, convert to string
+    for col in df_data.columns:
+        if df_data[col].dtype == 'object':
+            df_data[col] = df_data[col].astype(str)
+
+    text_gen = OpenAITextGenerator(
+        provider='openai',
+        api_type='azure',
+        azure_endpoint= os.getenv('AZURE_OPENAI_ENDPOINT'),
+        api_key= os.getenv('OPENAI_API_KEY'),
+        api_version = '2023-05-15',
+        
+    )
+    lida = Manager(text_gen=text_gen)
+
+    text_gen_config = TextGenerationConfig(
+        n = 1,
+        model = 'CapSuiteGPT35T16K',
+        temperature=0.1
+    )
+
+    summary = lida.summarize(df_data)
+    
+    print(f'*'*50)
+    goals = lida.goals(summary, n=1, textgen_config=text_gen_config,persona=f'Do not use white color for the line or bar.An data analyst of the company who want to know {question}')
+
+    # ValueError: Unsupported library. Choose from 'matplotlib', 'seaborn', 'plotly', 'bokeh', 'ggplot', 'altair'.
+
+    chart_result = []
+
+    final_explanation = []
+
+    for i in range(1):
+        try:
+            print(f"Goal{i}: {goals[i]}")
+            temp_chart = lida.visualize(summary=summary, goal=goals[i], textgen_config=text_gen_config,library='plotly')
+            temp_explanation = lida.explain(code=temp_chart[0].code)
+            final_explanation.append(temp_explanation)
+            chart_result.append(temp_chart)
+        except Exception as e:
+            print(f"Error while: {e}")
+    for i in range(len(chart_result)):
+        chart_result[i][0].savefig(f'chart_{i}.png')
+        print(f'*'*50)
+        print(f"Chart {i} saved")
+
+    # Path to your image
+    image_path = "chart_0.png"
+
+    # Open the image file
+    img = Image.open(image_path)
+    base64_image = encode_image(image_path)
+
+    llm = model
+
+    response = llm.invoke(
+        [
+            HumanMessage(
+                content=[
+                    {"type": "text", "text": "Give me some business insights base on the graph, contain exact number conclusion."},
+                    {
+                        "type": "image_url",
+                        "image_url": {
+                            "url": f"data:image/jpeg;base64,{base64_image}"
+                        },
+                    },
+                ]
+            )
+        ]
+    )
+    final_result_str = response.content
+
+    return final_result_str,img
+
+with gr.Blocks() as demo:
+    with gr.Row():
+        with gr.Column():
+            temp_img = gr.Image(
+                height=500
+            )
+        with gr.Column():
+            gr.ChatInterface(
+                    random_response,
+                    examples=['Top 10 prod_cate sales','Top product in category Seafood'],
+                    type="messages", 
+                    autofocus=False,
+                    additional_outputs=[temp_img]
+            )
+   
+demo.launch()

requirements.txt

@@ -0,0 +1,8 @@
+huggingface_hub==0.25.2
+sqlalchemy
+pandas
+lida
+llmx
+langchain_openai
+langchain_core
+gradio==5.11.0