updated app

app.py

@@ -77,11 +77,11 @@ def load_response_data() -> Dict[str, pd.DataFrame]:
 def get_questions_for_domain(domain):
     """Get list of questions for selected domain."""
     if "responses" not in DB or domain not in DB["responses"]:
-        return []
+        return gr.update(choices=[], value=None)
     
     df = DB["responses"][domain]
     questions = df['question'].unique().tolist()
-    return questions
+    return gr.update(choices=questions, value=None)
 
 def get_response_details(domain, question):
     """Get LLM answer, gold answer, and metrics for selected question."""
@@ -139,28 +139,63 @@ def get_dataset_choices():
         return []
 
 def get_data_preview():
-    """Returns separate dataframes for each domain."""
+    """Returns separate dataframes for each domain with columns reordered by type."""
     if "data" not in DB:
-        return {}, {}, {}, {}
+        return {}, {}, {}, {}, {}
     
     df = DB["data"]
     # Remove failed_samples column if it exists
     if 'failed_samples' in df.columns:
         df = df.drop(columns=['failed_samples'])
     
-    datasets = df['dataset_name'].unique()
+    # Define explicit domain order matching the UI
+    domain_order = ['pubmedqa', 'techqa', 'finqa', 'msmarco', 'cuad']
     
-    # Create separate dataframes for each domain
-    results = {}
-    for ds in datasets:
-        results[ds] = df[df['dataset_name'] == ds]
+    # Metric columns (Results)
+    result_cols = ['rmse_relevance', 'rmse_utilization', 'rmse_completeness', 'f1_score', 'aucroc']
+    metadata_cols = ['test_id', 'config_purpose', 'dataset_name']
     
-    # Return up to 4 domains (adjust if you have more)
-    domain_dfs = list(results.values())
-    while len(domain_dfs) < 4:
-        domain_dfs.append(pd.DataFrame())
+    domain_dfs = []
+    for ds in domain_order:
+        domain_df = df[df['dataset_name'] == ds].copy()
+        
+        if domain_df.empty:
+            domain_dfs.append(pd.DataFrame())
+            continue
+        
+        # Analyze constants and variables
+        consts, variables = analyze_domain_configs(domain_df)
+        
+        # Reorder columns: Metadata -> Constants -> Variables -> Results
+        ordered_cols = []
+        
+        # Add metadata columns first
+        for col in metadata_cols:
+            if col in domain_df.columns:
+                ordered_cols.append(col)
+        
+        # Add constant columns (sorted)
+        const_cols = sorted([col for col in consts.keys() if col in domain_df.columns])
+        ordered_cols.extend(const_cols)
+        
+        # Add variable columns (sorted)
+        var_cols = sorted([col for col in variables if col in domain_df.columns])
+        ordered_cols.extend(var_cols)
+        
+        # Add result columns
+        for col in result_cols:
+            if col in domain_df.columns:
+                ordered_cols.append(col)
+        
+        # Add any remaining columns
+        remaining = [col for col in domain_df.columns if col not in ordered_cols]
+        ordered_cols.extend(remaining)
+        
+        # Reorder dataframe
+        domain_df = domain_df[ordered_cols]
+        domain_dfs.append(domain_df)
     
-    return domain_dfs[0], domain_dfs[1], domain_dfs[2], domain_dfs[3]
+    return domain_dfs[0], domain_dfs[1], domain_dfs[2], domain_dfs[3], domain_dfs[4]
 
 def get_domain_state(dataset):
     empty_update = gr.update(visible=False, value=None, choices=[])
@@ -433,14 +468,28 @@ with gr.Blocks(title="RAG Analytics Pro") as demo:
         # TAB 2: Data Inspector
         with gr.TabItem("Data Preview"):
             gr.Markdown("### All Test Configurations by Domain")
+            
+            # Add column type legend
+            gr.Markdown("""
+            **Column Organization Guide:**
+            - **Metadata** (Test ID, Config Purpose, Dataset Name) - Test identifiers
+            - **Constants** - Same value across all tests in this domain (e.g., embedding model)
+            - **Variables** - Experimental parameters that change across tests (e.g., chunk size, reranker)
+            - **Results** - Performance metrics (RMSE Relevance, RMSE Utilization, RMSE Completeness, F1-Score, AUC-ROC)
+            
+            *Columns are automatically reordered: Metadata → Constants → Variables → Results*
+            """)
+            
             gr.Markdown("**Biomedical (PubMedQA)**")
             preview_table_1 = gr.Dataframe(interactive=False, wrap=True)
-            gr.Markdown("**Finance (FinQA)**")
+            gr.Markdown("**Customer Support (TechQA)**")
             preview_table_2 = gr.Dataframe(interactive=False, wrap=True)
-            gr.Markdown("**General (MS MARCO)**")
+            gr.Markdown("**Finance (FinQA)**")
             preview_table_3 = gr.Dataframe(interactive=False, wrap=True)
-            gr.Markdown("**Legal (CUAD)**")
+            gr.Markdown("**General (MS MARCO)**")
             preview_table_4 = gr.Dataframe(interactive=False, wrap=True)
+            gr.Markdown("**Legal (CUAD)**")
+            preview_table_5 = gr.Dataframe(interactive=False, wrap=True)
             preview_btn = gr.Button("Refresh Data Preview")
 
         # TAB 3: Comparison
@@ -534,7 +583,7 @@ with gr.Blocks(title="RAG Analytics Pro") as demo:
     )
     
     # Debug Preview Events
-    preview_btn.click(get_data_preview, inputs=None, outputs=[preview_table_1, preview_table_2, preview_table_3, preview_table_4])
+    preview_btn.click(get_data_preview, inputs=None, outputs=[preview_table_1, preview_table_2, preview_table_3, preview_table_4, preview_table_5])
     
     refresh_btn.click(
         generate_inter_domain_comparison,