Update app.py

app.py

@@ -3,161 +3,156 @@ import pandas as pd
 import os
 import matplotlib.pyplot as plt
 import seaborn as sns
-from sklearn.cluster import KMeans
-from langchain_openai import ChatOpenAI, OpenAIEmbeddings
-from langchain.chains import LLMChain
-from langchain.prompts import PromptTemplate
 from dotenv import load_dotenv
-from pycaret.regression import setup as setup_reg, compare_models as compare_reg, predict_model
-from pandasai import SmartDataframe
-from pandasai.llm.openai import OpenAI
-from langchain_community.document_loaders import UnstructuredExcelLoader, CSVLoader
+from io import BytesIO
+
+# AI & ML Components
+from langchain_openai import ChatOpenAI
+from pandasai import SmartDatalake
+from pandasai.llm import OpenAI
+from pycaret.regression import setup as setup_reg, compare_models, predict_model, pull, save_model, load_model
 from langchain_community.vectorstores import Chroma
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 
-# ✅ Fix: Set Streamlit Page Config at the Beginning (Before Any Other Streamlit Command)
-st.set_page_config(page_title="AI Store Replacement & Demand Prediction", layout="wide")
-
 # ✅ Load Environment Variables
 load_dotenv()
 openai_api_key = os.getenv("OPENAI_API_KEY")
 
 if not openai_api_key:
-    st.error("❌ OpenAI API Key is missing! Set it in your environment variables.")
-
-# ✅ Set chart save path
-chart_path = "/home/user/app/exports/charts/temp_chart.png"
-
-# ✅ Load KPI Excel File
-folder_path = os.getcwd()
-kpi_file = os.path.join(folder_path, "kpi.xlsx")
-st.info("📂 Loading KPI data...")
-try:
-    df_kpi = pd.read_excel(kpi_file, engine="openpyxl")
-    st.success(f"✅ KPI Data Loaded: {df_kpi.shape[0]} rows, {df_kpi.shape[1]} columns")
-except Exception as e:
-    st.error(f"❌ Error loading KPI data: {e}")
-
-# ✅ Show a preview of 5 unique rows instead of full dataset
-st.subheader("📊 Sample Data Preview (5 Unique Rows)")
-st.dataframe(df_kpi.drop_duplicates().sample(n=5, random_state=42) if len(df_kpi) > 5 else df_kpi)
-
-# ✅ Load Competitor Data with Error Handling
-competitor_files = {
-    "DEIRA CITY CENTRE": "Competitor Mapping Doc Final1(DEIRA CITY CENTRE).csv",
-    "ARABIAN CENTRE": "Competitor Mapping Doc Final1(ARABIAN CENTRE).csv",
-    "AL GHURAIR CENTRE": "Competitor Mapping Doc Final1(AL GHURAIR CENTRE).csv",
-    "IBN BATTUTA MALL": "Competitor Mapping Doc Final1(IBN BATTUTA MALL).csv",
-    "DUBAI HILLS": "Competitor Mapping Doc Final1(DUBAI HILLS).csv"
-}
-
-df_competitors = {}
-st.info("📂 Loading Competitor Data...")
-for mall, file_name in competitor_files.items():
-    file_path = os.path.join(folder_path, file_name)
-    if os.path.exists(file_path):
-        try:
-            df_competitors[mall] = pd.read_csv(file_path, encoding="ISO-8859-1")
-            st.success(f"✅ Competitor Data Loaded for {mall}: {df_competitors[mall].shape[0]} rows")
-        except Exception as e:
-            st.error(f"❌ Error loading {file_name}: {e}")
-
-# ✅ Load Data into ChromaDB Vector Store with File Error Handling
-st.info("🔄 Training AI embeddings on KPI and Competitor data...")
-vectorstore = None  # Initialize vectorstore to prevent NameError
-
-try:
-    loader_kpi = UnstructuredExcelLoader(kpi_file, mode="elements")
-    kpi_docs = loader_kpi.load()
-
-    competitor_docs = []
-    for mall, file_path in competitor_files.items():
-        if os.path.exists(file_path):
-            try:
-                loader = CSVLoader(file_path=file_path)
-                competitor_docs.extend(loader.load())
-            except Exception as e:
-                st.error(f"❌ Error processing {file_path} for embeddings: {e}")
-
-    all_docs = kpi_docs + competitor_docs
-
-    if len(all_docs) > 0:
-        # ✅ Use RecursiveCharacterTextSplitter for better chunking
-        text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
-        split_docs = text_splitter.split_documents(all_docs)
-
-        # ✅ Store in ChromaDB
-        embeddings = OpenAIEmbeddings(api_key=openai_api_key)
-        vectorstore = Chroma.from_documents(split_docs, embeddings)
-        st.success("✅ AI embeddings trained and stored in ChromaDB successfully!")
-    else:
-        st.warning("⚠️ No valid documents were found to train embeddings.")
-except Exception as e:
-    st.error(f"❌ Error training AI embeddings with ChromaDB: {e}")
-
-# ✅ Initialize Pandas AI Agent (SmartDataframe)
-st.info("🔧 Initializing Pandas AI Agent...")
-try:
-    llm = OpenAI(api_token=openai_api_key)
-    df_smart = SmartDataframe(df_kpi, config={"llm": llm})
-    st.success("✅ Pandas AI Agent Initialized!")
-except Exception as e:
-    st.error(f"❌ Error initializing Pandas AI Agent: {e}")
+    st.error("❌ OpenAI API Key is missing! Set it in your .env file.")
+
+# ✅ Streamlit Page Configuration
+st.set_page_config(page_title="AI-Powered Store Recommender & Chatbot", layout="wide")
 
 # ✅ Title
-st.title("🔄 AI-Powered Store Replacement & Forecasting")
+st.title("🛍️ AI-Powered Apparel Group Store Recommender & Chatbot")
 
-# ✅ Store Replacement System
-st.subheader("🏬 AI-Based Store Replacement")
+# 📂 **File Upload Section**
+uploaded_kpi = st.file_uploader("📂 Upload KPI Excel file (.xlsx)", type=["xlsx"], key="file_uploader_kpi")
+uploaded_competitor = st.file_uploader("📂 Upload Competitor CSV file (.csv)", type=["csv"], key="file_uploader_competitor")
 
-selected_mall = st.selectbox("🏬 Select a Mall:", df_kpi["Mall"].unique())
+if uploaded_kpi and uploaded_competitor:
+    # ✅ Load KPI Data
+    df_kpi = pd.read_excel(uploaded_kpi, engine="openpyxl")
+    df_competitor = pd.read_csv(uploaded_competitor, encoding="utf-8", on_bad_lines="skip")
 
-if selected_mall:
-    st.info(f"📌 Selected Mall: {selected_mall}")
-    mall_data = df_kpi[df_kpi["Mall"] == selected_mall]
+    # ✅ Display Previews
+    st.subheader("📊 KPI Data Preview (5 Unique Rows)")
+    st.dataframe(df_kpi.drop_duplicates().sample(n=5, random_state=42) if len(df_kpi) > 5 else df_kpi)
 
-    store_to_replace = st.selectbox("🚪 Select the store that is closing:", mall_data["Store_Brand"].unique())
+    st.subheader("🏆 Competitor Data Preview (5 Unique Rows)")
+    st.dataframe(df_competitor.drop_duplicates().sample(n=5, random_state=42) if len(df_competitor) > 5 else df_competitor)
 
-    if store_to_replace:
-        st.info(f"📌 Store to Replace: {store_to_replace}")
+    # ✅ Select Mall for Competitor Insights
+    st.subheader("🔎 Competitor Analysis")
+
+    selected_mall = st.selectbox("🏬 Select a Mall for Competitor Insights:", df_kpi["Mall"].unique())
+
+    if selected_mall:
+        competitors = df_competitor[df_competitor["Mall"] == selected_mall]["Store_Brand"].unique()
+
+        if len(competitors) > 0:
+            st.write(f"💡 **Competitors in {selected_mall}:**")
+            st.write(competitors)
+        else:
+            # ✅ Use AI to Predict Competitors for Malls Not in Dataset
+            llm = ChatOpenAI(model="gpt-4", api_key=openai_api_key)
+            prompt = f"Based on the given KPI and competitor data, suggest competitors for stores in {selected_mall}."
+            ai_competitors = llm.invoke(prompt)
 
-        llm = ChatOpenAI(model="gpt-4", api_key=openai_api_key)
+            st.warning(f"⚠️ No direct competitor data found. AI has predicted competitors:")
+            st.write(ai_competitors)
 
+    # ✅ AI-Generated Store Recommendations
+    st.subheader("🔮 AI-Based Store Recommendations")
+
+    store_to_replace = st.selectbox("🚪 Select a Store to Replace:", df_kpi["Store_Brand"].unique())
+
+    if store_to_replace:
         # ✅ Identify Store Category
-        category_prompt = PromptTemplate.from_template(
-            "What category does {store_name} belong to? Choose from: Fashion, Electronics, Footwear, Food, Home Decor, etc."
-        )
-        category_chain = LLMChain(llm=llm, prompt=category_prompt)
-        store_category = category_chain.run(store_name=store_to_replace)
+        prompt = f"Identify the category of {store_to_replace} (e.g., Fashion, Electronics, Footwear, etc.)."
+        store_category = llm.invoke(prompt)
+
         st.write(f"🔍 **Identified Category:** {store_category}")
 
-        # ✅ Find Competitor Stores
-        if selected_mall in df_competitors:
-            competitor_list = df_competitors[selected_mall]["Competitor_Stores"].dropna().tolist()
-            st.success(f"✅ Retrieved {len(competitor_list)} competitors from dataset.")
+        # ✅ Find 3 AI-Recommended Replacement Brands NOT Present in This Mall
+        all_brands = set(df_kpi["Store_Brand"].unique())
+        existing_brands = set(df_kpi[df_kpi["Mall"] == selected_mall]["Store_Brand"].unique())
+
+        available_replacements = list(all_brands - existing_brands)
+
+        if available_replacements:
+            prompt = f"Suggest 3 alternative brands in {store_category} category that are not present in {selected_mall}. Possible replacements: {available_replacements}"
+            ai_recommendations = llm.invoke(prompt)
+
+            st.success(f"🔄 **AI Recommended Replacements for {store_to_replace}:**")
+            st.write(ai_recommendations)
         else:
-            st.warning("⚠️ No competitor data found. Using AI embeddings to predict competitors.")
-            if vectorstore:
-                search_results = vectorstore.similarity_search(f"Find competitors for {store_category} in {selected_mall}")
-                competitor_list = [doc.page_content for doc in search_results]
-            else:
-                competitor_list = ["No AI embeddings available."]
-
-        st.subheader("🔎 Competitors in This Mall")
-        for competitor in competitor_list:
-            st.write(f"- {competitor}")
-
-# ✅ AI Chatbot (Using PandasAI + ChromaDB)
-st.subheader("💬 AI-Powered Business Chatbot")
-user_query = st.text_input("🔍 Ask AI About Your Data:")
-
-if user_query:
-    st.info(f"🧠 Processing AI Query: {user_query}")
-    if "chart" in user_query.lower():
-        st.image(chart_path, caption="Generated Chart", use_column_width=True)
-    else:
-        ai_response = df_smart.chat(user_query)
-        st.markdown(f"🤖 **AI Response:** {ai_response}")
+            st.warning("⚠️ No new brands found for replacement.")
+
+    # ✅ Machine Learning-Based Sales Prediction
+    st.subheader("📈 Predict Annual Sales for a New Store")
+
+    selected_brand = st.selectbox("🛍️ Select a Brand:", df_kpi["Store_Brand"].unique())
+
+    if selected_brand:
+        # **Step 1: Select Features**
+        if "Annual_Sales" in df_kpi.columns:
+            numeric_cols = df_kpi.select_dtypes(include=["number"]).columns.tolist()
+            feature_cols = st.multiselect("📊 Select Features for Prediction:", numeric_cols, default=numeric_cols)
+
+            if feature_cols:
+                # **Step 2: Train Model (Only Once Per File)**
+                model_save_path = "best_sales_model"
+                if not os.path.exists(model_save_path):
+                    st.info("🔄 Training the ML Model... (This may take a few seconds)")
+                    setup_reg(df_kpi[feature_cols + ["Annual_Sales"]], target="Annual_Sales", silent=True, session_id=123)
+                    best_model = compare_models()
+                    save_model(best_model, model_save_path)
+                    st.success("✅ Model Training Complete! Best model saved.")
+
+                # **Step 3: Load Trained Model**
+                best_model = load_model(model_save_path)
+
+                # **Step 4: Predict Sales for a New Store**
+                ref_store = df_kpi[df_kpi["Store_Brand"] == selected_brand]
+                
+                if not ref_store.empty:
+                    new_store_data = ref_store[feature_cols].mean().to_frame().T  # Use mean as input
+                    predicted_sales = predict_model(best_model, data=new_store_data)["Label"].iloc[0]
+
+                    st.write(f"📊 **Predicted Annual Sales for {selected_brand}:** **${predicted_sales:,.2f}**")
+
+                    # **Feature Importance Chart**
+                    st.subheader("🔍 Key Factors Affecting Sales")
+                    feature_importance = pull()
+                    st.bar_chart(feature_importance)
+
+                else:
+                    st.warning(f"⚠️ No reference data available to predict sales for {selected_brand}.")
+
+    # ✅ AI Chatbot for Business Insights
+    st.subheader("💬 AI Chatbot for Business Insights")
+
+    # ✅ Configure AI for PandasAI Queries
+    sdl = SmartDatalake([df_kpi, df_competitor], config={
+        "llm": OpenAI(api_token=openai_api_key),
+        "use_error_correction_framework": True,
+        "enable_cache": False,
+        "enable_restricted_mode": False,
+        "save_charts": True,
+        "save_charts_path": "/home/user/app/exports/charts/temp_chart.png"
+    })
+
+    user_query = st.text_input("🔍 Ask AI About the Dataset:")
+    if user_query:
+        with st.spinner("🤖 AI is analyzing..."):
+            try:
+                ai_response = sdl.chat(user_query)
+                st.markdown(f"🤖 **AI Response:** {ai_response}")
+            except Exception as e:
+                st.error(f"❌ AI Error: {str(e)}")
 
+# ✅ Footer
 st.markdown("---")
-st.markdown("💡 Built with ❤️ using **ChromaDB, AI Store Replacement & Business Analytics**")
+st.markdown("💡 Built with ❤️ using **Streamlit, PandasAI, PyCaret & GPT-4**")