Fix: use graph data from session_state when file system is unavailable (HF Spaces compatibility)

eda/pages/00_Graph_Generation.py

@@ -443,26 +443,44 @@ if selected_json:
     if not file_path.is_absolute():
         file_path = parent_dir / selected_json
     
-    # Check if file exists before accessing stats
-    if not file_path.exists():
+    # Check if file exists - if not, try to use session_state data
+    file_exists = file_path.exists()
+    
+    # If file doesn't exist but we have data in session_state (just generated), use that
+    use_session_data = False
+    if not file_exists and just_generated and 'pipeline_graph_json' in st.session_state:
+        graph_data = st.session_state['pipeline_graph_json']['data']
+        use_session_data = True
+        st.info("📦 Using graph data from current session (file system is temporary on HF Spaces)")
+    elif not file_exists:
         st.error(f"❌ File not found: `{file_path.name}`")
         st.warning("The file may have been moved or renamed. Please refresh the page or select another file.")
         st.stop()
     
-    file_size = file_path.stat().st_size / 1024 / 1024
-    file_time = datetime.fromtimestamp(file_path.stat().st_mtime)
-    
-    # Load JSON to extract graph metadata
-    try:
-        with open(file_path, 'r', encoding='utf-8') as f:
-            graph_metadata = json.load(f)
-        num_nodes = len(graph_metadata.get('nodes', []))
-        num_links = len(graph_metadata.get('links', []))
-        model_id = graph_metadata.get('metadata', {}).get('model_id', 'N/A')
-    except Exception:
-        num_nodes = None
-        num_links = None
-        model_id = None
+    # Get file stats and metadata
+    if use_session_data:
+        # Use data from session_state
+        file_size = len(json.dumps(graph_data)) / 1024 / 1024  # Approximate size
+        file_time = datetime.fromisoformat(st.session_state['pipeline_graph_json']['timestamp'])
+        num_nodes = len(graph_data.get('nodes', []))
+        num_links = len(graph_data.get('links', []))
+        model_id = graph_data.get('metadata', {}).get('model_id', 'N/A')
+    else:
+        # Use file on disk
+        file_size = file_path.stat().st_size / 1024 / 1024
+        file_time = datetime.fromtimestamp(file_path.stat().st_mtime)
+        
+        # Load JSON to extract graph metadata
+        try:
+            with open(file_path, 'r', encoding='utf-8') as f:
+                graph_metadata = json.load(f)
+            num_nodes = len(graph_metadata.get('nodes', []))
+            num_links = len(graph_metadata.get('links', []))
+            model_id = graph_metadata.get('metadata', {}).get('model_id', 'N/A')
+        except Exception:
+            num_nodes = None
+            num_links = None
+            model_id = None
     
     # Display file info and graph metadata
     col1, col2, col3 = st.columns(3)
@@ -487,9 +505,15 @@ if selected_json:
     if st.button(button_label, key="extract_existing", type="primary"):
         try:
             with st.spinner("Extracting metrics..."):
-                json_full_path = str(parent_dir / selected_json)
-                with open(json_full_path, 'r', encoding='utf-8') as f:
-                    graph_data = json.load(f)
+                # Use graph_data from session_state if available, otherwise load from file
+                if use_session_data:
+                    # Already have graph_data from session_state
+                    pass
+                else:
+                    # Load from file
+                    json_full_path = str(parent_dir / selected_json)
+                    with open(json_full_path, 'r', encoding='utf-8') as f:
+                        graph_data = json.load(f)
                 
                 csv_output_path = str(parent_dir / "output" / "graph_feature_static_metrics.csv")
                 df = extract_static_metrics_from_json(