Update src/streamlit_app.py

src/streamlit_app.py

@@ -15,16 +15,12 @@ client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
 
 PROMPT_INSTRUCTIONS_TEXT = """
 You are a forensic auditor AI with deep domain expertise and a sharp eye for irregularities. Your job is to identify **anomalies** in a single column of financial data.
-
 Analyze the values provided and return only values that are:
 - **Numerical outliers**: extremely high/low or oddly rounded numbers
 - **Format inconsistencies**: strange symbols, irregular formatting, or data corruption
 - **Rare or suspicious values**: strings or categories that do not appear to fit the overall pattern
-
 ONLY analyze the values from the provided column, without relying on any external context.
-
 Return ONLY the following JSON object and nothing else:
-
 {
   "anomalies": [
     {
@@ -91,27 +87,31 @@ st.markdown("""
 This tool combines machine learning and large language models to detect anomalies in datasets. We first apply isolation forest to the full dataset to flag data-level outliers. Then, you can select one column to perform a second pass of analysis using OpenAI's GPT-4, which focuses on semantic and contextual anomalies within that column only (e.g. Payment_Method column).
 """)
 
-# Button to load sample data
-df = None
-sample_loaded = False
+# Initialize session state for df
+if "df" not in st.session_state:
+    st.session_state.df = None
+
+# Load sample data
 if st.button("Load sample dataset"):
-    sample_path = "src/df_crypto.csv"
     try:
-        df = pd.read_csv(sample_path)
-        sample_loaded = True
+        st.session_state.df = pd.read_csv("src/df_crypto.csv")
         st.success("Sample dataset loaded from `src/df_crypto.csv`.")
     except Exception as e:
         st.error(f"Could not load sample dataset: {e}")
 
 # File upload
-if not sample_loaded:
+if st.session_state.df is None:
     uploaded_file = st.file_uploader("Or upload your own CSV file", type=["csv"])
     if uploaded_file:
         try:
-            df = pd.read_csv(uploaded_file)
+            st.session_state.df = pd.read_csv(uploaded_file)
+            st.success("Custom dataset uploaded.")
         except Exception as e:
             st.error(f"Could not read uploaded CSV. Error: {e}")
 
+# Use persisted df
+df = st.session_state.df
+
 if df is not None:
     st.subheader("Full Dataset")
     st.dataframe(df, use_container_width=True)
@@ -126,7 +126,6 @@ if df is not None:
 
     # ---------------- LLM Section ----------------
     st.markdown("### LLM-Based Anomaly Detection (specific column)")
-
     selected_column = st.selectbox("Select a column to analyze with LLM:", df.columns)
 
     if st.button("Run LLM Anomaly Detection on selected column"):