Add SCF convergence check and improve cube file parsing robustness

app.py

@@ -4,41 +4,66 @@ import numpy as np
 
 def parse_cube_file(cube_file):
     """Parse a cube file and return grid coordinates and values."""
-    with open(cube_file, 'r') as f:
-        lines = f.readlines()
-    
-    # Skip comments
-    natoms = int(lines[2].split()[0])
-    origin = np.array([float(x) for x in lines[2].split()[1:4]])
-    
-    # Skip atom lines
-    grid_line = lines[3 + natoms]
-    nx = int(grid_line.split()[0])
-    dx = float(grid_line.split()[1])
-    
-    grid_line = lines[4 + natoms]
-    ny = int(grid_line.split()[1])
-    dy = float(grid_line.split()[2])
-    
-    grid_line = lines[5 + natoms]
-    nz = int(grid_line.split()[2])
-    dz = float(grid_line.split()[3])
-    
-    # Read data
-    data_start = 6 + natoms
-    data = []
-    for line in lines[data_start:]:
-        data.extend([float(x) for x in line.split()])
-    
-    # Reshape data
-    values = np.array(data).reshape((nx, ny, nz))
-    
-    # Create coordinate grids
-    x = origin[0] + np.arange(nx) * dx
-    y = origin[1] + np.arange(ny) * dy
-    z = origin[2] + np.arange(nz) * dz
-    
-    return x, y, z, values
+    try:
+        with open(cube_file, 'r') as f:
+            lines = f.readlines()
+        
+        if len(lines) < 6:
+            raise ValueError("Cube file too short")
+        
+        # Find the line with natoms (should be the third line or so)
+        for i, line in enumerate(lines):
+            parts = line.split()
+            if len(parts) >= 4:
+                try:
+                    natoms = int(float(parts[0]))  # Try to parse as int, allowing for float strings
+                    if natoms > 0:
+                        break
+                except ValueError:
+                    continue
+        else:
+            raise ValueError("Could not find natoms in cube file")
+        
+        # Now i is the line with natoms
+        origin = np.array([float(x) for x in lines[i].split()[1:4]])
+        
+        # Skip atom lines
+        grid_start = i + 1 + natoms
+        if grid_start >= len(lines):
+            raise ValueError("Cube file malformed")
+        
+        grid_line = lines[grid_start]
+        nx = int(grid_line.split()[0])
+        dx = float(grid_line.split()[1])
+        
+        grid_line = lines[grid_start + 1]
+        ny = int(grid_line.split()[1])
+        dy = float(grid_line.split()[2])
+        
+        grid_line = lines[grid_start + 2]
+        nz = int(grid_line.split()[2])
+        dz = float(grid_line.split()[3])
+        
+        # Read data
+        data_start = grid_start + 3
+        data = []
+        for line in lines[data_start:]:
+            data.extend([float(x) for x in line.split()])
+        
+        if len(data) != nx * ny * nz:
+            raise ValueError(f"Data size mismatch: expected {nx*ny*nz}, got {len(data)}")
+        
+        # Reshape data
+        values = np.array(data).reshape((nx, ny, nz))
+        
+        # Create coordinate grids
+        x = origin[0] + np.arange(nx) * dx
+        y = origin[1] + np.arange(ny) * dy
+        z = origin[2] + np.arange(nz) * dz
+        
+        return x, y, z, values
+    except Exception as e:
+        raise ValueError(f"Error parsing cube file: {e}")
 
 
 import gradio as gr
@@ -195,7 +220,10 @@ def smiles_to_molecular_orbitals(smiles_input: str, name_input: str) -> str:
 
             # Run Hartree-Fock calculation
             mf = pyscf.scf.RHF(mol_pyscf)
-            mf.kernel()
+            energy = mf.kernel()
+            
+            if not mf.converged:
+                return "<p>Hartree-Fock calculation did not converge. Try a smaller molecule or different geometry.</p>"
 
             # Get HOMO and LUMO indices
             nocc = mol_pyscf.nelectron // 2

force_rebuild.txt

@@ -1,2 +1,2 @@
 # Force rebuild
-2025-11-08 v5
+2025-11-08 v6