fix: Move heavy imports to runtime in spatial_genes.py

Moved torch, numpy, pandas, and sklearn imports from module level to
runtime within functions to avoid import errors in minimal environments.
This allows the CLI to start without requiring all optional dependencies.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: cafferychen777

chatspatial/tools/spatial_genes.py

@@ -9,31 +9,25 @@
 import sys
 import tempfile
 import shutil
-import numpy as np
-import pandas as pd
-import torch
-import torch.nn as nn
-from typing import Dict, List, Tuple, Optional, Any
+from typing import Dict, List, Tuple, Optional, Any, TYPE_CHECKING
 from pathlib import Path
-import scanpy as sc
-from sklearn.decomposition import PCA
-from sklearn.preprocessing import StandardScaler
 import warnings
 import logging
 
+if TYPE_CHECKING:
+    import numpy as np
+    import pandas as pd
+    import torch
+    import torch.nn as nn
+    import scanpy as sc
+    from sklearn.decomposition import PCA
+    from sklearn.preprocessing import StandardScaler
+
 logger = logging.getLogger(__name__)
 
-# Try to import GASTON from standard Python package installation
-try:
-    import gaston
-    from gaston import neural_net, spatial_gene_classification, binning_and_plotting
-    from gaston import dp_related, segmented_fit, process_NN_output
-    GASTON_AVAILABLE = True
-    GASTON_IMPORT_ERROR = None
-except ImportError as e:
-    GASTON_AVAILABLE = False
-    # Only show warning when GASTON is actually requested
-    GASTON_IMPORT_ERROR = str(e)
+# GASTON import will be done at runtime
+GASTON_AVAILABLE = None
+GASTON_IMPORT_ERROR = None
 
 from ..models.data import SpatialVariableGenesParameters
 from ..models.analysis import SpatialVariableGenesResult
@@ -98,6 +92,25 @@ async def _identify_spatial_genes_gaston(
     context=None
 ) -> SpatialVariableGenesResult:
     """Identify spatial variable genes using GASTON method."""
+    # Import dependencies at runtime
+    import numpy as np
+    import pandas as pd
+    import torch
+    import torch.nn as nn
+    
+    # Check GASTON availability at runtime
+    global GASTON_AVAILABLE, GASTON_IMPORT_ERROR
+    if GASTON_AVAILABLE is None:
+        try:
+            import gaston
+            from gaston import neural_net, spatial_gene_classification, binning_and_plotting
+            from gaston import dp_related, segmented_fit, process_NN_output
+            GASTON_AVAILABLE = True
+            GASTON_IMPORT_ERROR = None
+        except ImportError as e:
+            GASTON_AVAILABLE = False
+            GASTON_IMPORT_ERROR = str(e)
+    
     if not GASTON_AVAILABLE:
         error_msg = (
             f"GASTON is not available: {GASTON_IMPORT_ERROR}\n\n"
@@ -231,6 +244,10 @@ async def _identify_spatial_genes_spatialde(
     context=None
 ) -> SpatialVariableGenesResult:
     """Identify spatial variable genes using SpatialDE method."""
+    # Import dependencies at runtime
+    import numpy as np
+    import pandas as pd
+    import scanpy as sc
     try:
         import SpatialDE
         from SpatialDE.util import qvalue
@@ -355,7 +372,9 @@ async def _identify_spatial_genes_spatialde(
     return result
 
 
-async def _gaston_feature_engineering_glmpca(adata, n_components: int, context) -> np.ndarray:
+async def _gaston_feature_engineering_glmpca(adata, n_components: int, context):
+    # Import dependencies at runtime
+    import numpy as np
     """GASTON-specific feature engineering using GLM-PCA (algorithm requirement)."""
     try:
         from glmpca.glmpca import glmpca
@@ -387,7 +406,11 @@ async def _gaston_feature_engineering_glmpca(adata, n_components: int, context)
     return glmpca_result["factors"]
 
 
-async def _gaston_feature_engineering_pearson(adata, n_components: int, context) -> np.ndarray:
+async def _gaston_feature_engineering_pearson(adata, n_components: int, context):
+    # Import dependencies at runtime
+    import numpy as np
+    import scanpy as sc
+    from sklearn.decomposition import PCA
     """GASTON-specific feature engineering using Pearson residuals PCA (algorithm requirement)."""
     if context:
         await context.info("Computing GASTON Pearson residuals feature engineering (algorithm requirement)")
@@ -409,6 +432,10 @@ async def _train_gaston_model(
     context
 ) -> Tuple[Any, List[float], float]:
     """Train GASTON neural network model."""
+    # Import dependencies at runtime
+    import numpy as np
+    import torch
+    import torch.nn as nn
 
     # Load data
     S = np.load(coords_file)
@@ -442,11 +469,15 @@ async def _train_gaston_model(
 
 
 async def _analyze_spatial_patterns(
-    model, spatial_coords: np.ndarray, expression_features: np.ndarray,
+    model, spatial_coords, expression_features,
     adata, params: SpatialVariableGenesParameters, context
 ) -> Dict[str, Any]:
     """Analyze spatial patterns from trained GASTON model using complete GASTON workflow."""
-
+    # Import dependencies at runtime
+    import numpy as np
+    import pandas as pd
+    import torch
+    
     if context:
         await context.info("Processing neural network output following GASTON tutorial")
 
@@ -649,6 +680,9 @@ async def _identify_spatial_genes_spark(
     context=None
 ) -> SpatialVariableGenesResult:
     """Identify spatial variable genes using SPARK method."""
+    # Import dependencies at runtime
+    import numpy as np
+    import pandas as pd
     try:
         from rpy2 import robjects as ro
         from rpy2.robjects import conversion, default_converter
@@ -854,6 +888,8 @@ async def _identify_spatial_genes_spark(
 
 def _set_random_seeds(seed: int):
     """Set random seeds for reproducibility."""
+    import numpy as np
+    import torch
     np.random.seed(seed)
     torch.manual_seed(seed)
     if torch.cuda.is_available():