feat: train all models on full data and save their paths

app/training/train_classifier.py

@@ -178,9 +178,23 @@ def train(
         title=f"Best: {best_name}",
     )
 
-    # ── Train final model on ALL data ──────────────
-    print(f"\nTraining final {best_name} on all {len(y)} samples...")
-    best_model.fit(X_scaled, y)
+    # ── Train ALL models on full data ────────────────
+    all_model_paths: dict[str, str] = {}
+    for name, model in candidates:
+        print(f"\nTraining final {name} on all {len(y)} samples...")
+        model.fit(X_scaled, y)
+        safe_name = name.lower().replace(" ", "_").replace("(", "").replace(")", "")
+        model_pkl = models_dir / f"model_{safe_name}.pkl"
+        with open(model_pkl, "wb") as f:
+            pickle.dump(model, f)
+        all_model_paths[name] = str(model_pkl)
+        print(f"  → Saved: {model_pkl}")
+
+    best_model = None
+    for name, model in candidates:
+        if name == best_name:
+            best_model = model
+            break
 
     # ── Feature importance ─────────────────────────
     importance_data = _extract_importance(best_model, feature_cols)
@@ -215,6 +229,7 @@ def train(
     json_results["_n_features"] = X.shape[1]
     json_results["_n_folds"] = n_folds
     json_results["_data_leakage_fix"] = "duration_sec and sample_rate removed from features (v2)"
+    json_results["_model_paths"] = all_model_paths
     if importance_data:
         json_results["_feature_importance"] = {
             name: round(imp, 6) for name, imp in importance_data