Update preprocessing.py

preprocessing.py

@@ -161,14 +161,25 @@ def get_latest_sequence(
         "electricity_rate",    # 14
     ]
 
-    df_miner = df_miner.dropna().reset_index(drop=True)
+    # ---------------------------------------------------------
+    # Ensure we only use rows with complete core blockchain data.
+    # This avoids including a "today" row where, e.g., difficulty or
+    # hashrate are still NaN while price is already updated.
+    # ---------------------------------------------------------
+    df_miner = df_miner.sort_values("date").reset_index(drop=True)
+
+    core_cols = ["bitcoin_price", "difficulty", "fees", "hashrate", "revenue"]
+    df_miner = df_miner.dropna(subset=core_cols)
 
     if len(df_miner) < window_size:
         raise ValueError(
-            f"Not enough data to build a {window_size}-day window, got {len(df_miner)} rows."
+            f"Not enough data to build a {window_size}-day window after dropping NaNs. "
+            f"Have {len(df_miner)} rows, need at least {window_size}."
         )
 
+    # Take the last `window_size` fully-populated days
     df_window = df_miner.tail(window_size).reset_index(drop=True)
+
     sequence = df_window[feature_cols].values.astype(float)
     pred_date = df_window["date"].iloc[-1]