fix: robust live prediction and playground UI cleanup

webapp/main.py

@@ -191,6 +191,9 @@ async def benchmark(
                 feature_types[col] = "categorical"
         result["dataset_info"]["feature_types"] = feature_types
         
+        # Add a sample row for the playground preview
+        result["dataset_info"]["preview"] = [df.head(1).fillna("").to_dict("records")[0]]
+        
         # Cache the Best Overall model for the Live Playground
         best_name = result["recommendation"]["recommendations"]["best_overall"]["model"]
         X = df.drop(columns=[target_col])
@@ -243,12 +246,17 @@ async def predict(data: dict):
     try:
         # Convert input dict to DataFrame
         input_df = pd.DataFrame([data])
-        # Ensure column order matches training
+        # Ensure column order matches training, filling missing with 0/empty
+        for col in CHAMPION_INFO["features"]:
+            if col not in input_df.columns:
+                input_df[col] = 0
         input_df = input_df[CHAMPION_INFO["features"]]
         
         # Use the EXACT same encoders that were used during training
         X_test, _ = _prep(input_df, encoders=CHAMPION_INFO.get("encoders"))
         
+        logger.info(f"Predicting for {CHAMPION_INFO['name']}...")
+        
         if CHAMPION_INFO["task"] == "classification":
             raw_pred = np.array(CHAMPION_MODEL.predict(X_test))
             # Flatten if nested (CatBoost/Sklearn sometimes return [[val]] or [val])

webapp/static/app.js

@@ -502,7 +502,8 @@ function renderPlayground(datasetInfo, bestOverall, task) {
   if (!form || !bestOverall) return;
   form.innerHTML = "";
   
-  const features = datasetInfo.columns || [];
+  const targetCol = datasetInfo.target_col;
+  const features = (datasetInfo.columns || []).filter(c => c !== targetCol);
   const preview = datasetInfo.preview ? datasetInfo.preview[0] : {};
 
   features.forEach(f => {