Update core/infer.py

core/infer.py

@@ -1,25 +1,38 @@
 import numpy as np, pandas as pd
 import shap
+import os, importlib.util
 
-# --- import that works both as package and script ---
-try:
-    from .features import build_features
-    from .normalize import robust_industry_z
-    from .external_score import (
-        get_external_template_df,
-        apply_llm_signals_to_df,
-        score_external_from_df,
-    )
-    from .ai_judgement import ai_evaluate, suggest_external_with_llm
-except ImportError:
-    from core.features import build_features
-    from core.normalize import robust_industry_z
-    from core.external_score import (
-        get_external_template_df,
-        apply_llm_signals_to_df,
-        score_external_from_df,
-    )
-    from core.ai_judgement import ai_evaluate, suggest_external_with_llm
+# --- robust import of sibling modules ---
+def _import_sibling(mod_name: str, filename: str):
+    try:
+        return __import__(f"{__package__}.{mod_name}", fromlist=[mod_name])  # relative
+    except Exception:
+        pass
+    try:
+        return __import__(f"core.{mod_name}", fromlist=[mod_name])          # absolute
+    except Exception:
+        pass
+    cur_dir = os.path.dirname(os.path.abspath(__file__))
+    path = os.path.join(cur_dir, filename)
+    spec = importlib.util.spec_from_file_location(mod_name, path)
+    if spec and spec.loader:
+        m = importlib.util.module_from_spec(spec)
+        spec.loader.exec_module(m)
+        return m
+    raise ImportError(f"Failed to import sibling module '{mod_name}'")
+
+_features = _import_sibling("features", "features.py")
+_normalize = _import_sibling("normalize", "normalize.py")
+_external = _import_sibling("external_score", "external_score.py")
+_ai = _import_sibling("ai_judgement", "ai_judgement.py")
+
+build_features = _features.build_features
+robust_industry_z = _normalize.robust_industry_z
+get_external_template_df = _external.get_external_template_df
+apply_llm_signals_to_df = _external.apply_llm_signals_to_df
+score_external_from_df = _external.score_external_from_df
+ai_evaluate = _ai.ai_evaluate
+suggest_external_with_llm = _ai.suggest_external_with_llm
 # ---------------------------------------------------
 
 Z_BASE = ["opm","npm","cur_ratio","eq_ratio","de","ocf_sales"]
@@ -72,6 +85,7 @@ class InferenceService:
             + self.weights["ai"]*ai["ai_total"]
         )
 
+        # SHAPの両対応
         try:
             sv = self.explainer.shap_values(df[self.feats])
             shap_vals = (sv[0][0] if isinstance(sv, list) else sv[0]).tolist()