fix data loading

data.py

@@ -57,7 +57,10 @@ KEYS_TO_KEEP = [
     "job_link_nvidia",
 ]
 
+# ============================================================================
 # HELPER FUNCTIONS
+# ============================================================================
+
 def generate_fake_dates(num_days: int = 7) -> List[str]:
     """Generate fake dates for the last N days."""
     today = datetime.now()
@@ -85,7 +88,10 @@ def safe_extract(row: pd.Series, key: str) -> int:
     """Safely extract an integer value from a DataFrame row."""
     return int(row.get(key, 0)) if pd.notna(row.get(key, 0)) else 0
 
+# ============================================================================
 # DATA LOADING FUNCTIONS
+# ============================================================================
+
 def log_dataframe_link(link: str) -> str:
     """
     Adds the link to the dataset in the logs, modifies it to get a clockable link and then returns the date of the 
@@ -164,11 +170,17 @@ def get_available_dates() -> List[str]:
         common_dates = sorted(amd_dates.intersection(nvidia_dates), reverse=True)
         logger.info(f"Common dates: {len(common_dates)} dates where both AMD and NVIDIA have data")
         
-        return common_dates[:30] if common_dates else generate_fake_dates()
+        if common_dates:
+            return common_dates[:30]  # Limit to last 30 days
+        
+        # No real dates available - log warning and return empty list
+        # This will allow the system to fall back to sample data properly
+        logger.warning("No common dates found between AMD and NVIDIA datasets")
+        return []
         
     except Exception as e:
         logger.error(f"Error getting available dates: {e}")
-        return generate_fake_dates()
+        return []
 
 
 def get_data_for_date(target_date: str) -> tuple[pd.DataFrame, str]:
@@ -501,7 +513,8 @@ class CIResults:
             self.sample_data = True
             new_df, latest_update_msg = get_sample_data()
             self.latest_update_msg = latest_update_msg
-            self.available_dates = None
+            # Generate fake dates for sample data historical functionality
+            self.available_dates = generate_fake_dates()
         
         # Update attributes
         self.df = new_df
@@ -531,8 +544,9 @@ class CIResults:
         """Load all available historical data at startup."""
         try:
             if not self.available_dates:
-                self.available_dates = generate_fake_dates()
-                logger.info(f"No available dates found, generated {len(self.available_dates)} sample dates.")
+                logger.warning("No available dates found, skipping historical data load")
+                self.all_historical_data = pd.DataFrame()
+                return
             
             logger.info(f"Loading all historical data for {len(self.available_dates)} dates...")
             start_date, end_date = self.available_dates[-1], self.available_dates[0]