fix

data.py

@@ -433,43 +433,34 @@ def find_failure_first_seen(historical_df: pd.DataFrame, model_name: str, test_n
 
 def find_new_regressions(current_df: pd.DataFrame, historical_df: pd.DataFrame) -> list[dict]:
     """
-    Compare current failures against the previous day's failures to find new regressions.
+    Compare CURRENT failures against PREVIOUS day's failures to find NEW regressions.
     
-    Returns a list of dicts with:
-    - model: model name
-    - test: test name
-    - device: 'amd' or 'nvidia'
-    - gpu_type: 'single' or 'multi'
+    A regression is a test that:
+    - Is failing in the CURRENT/LATEST run (current_df)
+    - Was NOT failing in the PREVIOUS run (yesterday in historical_df)
     """
     if current_df.empty or historical_df.empty:
         return []
     
     new_regressions = []
     
-    # Get the two most recent dates
+    # Get the most recent date from historical data (this is "yesterday")
     available_dates = sorted(historical_df['date'].unique(), reverse=True)
-    if len(available_dates) < 2:
-        # Not enough history to compare
+    if len(available_dates) < 1:
+        # No history to compare against
         return []
     
-    today_date = available_dates[0]
-    yesterday_date = available_dates[1]
-    
-    # Get data for both dates
-    today_data = historical_df[historical_df['date'] == today_date]
+    yesterday_date = available_dates[0]
     yesterday_data = historical_df[historical_df['date'] == yesterday_date]
     
-    # For each model, compare failures
+    # For each model in current data, compare against yesterday
     for model_name in current_df.index:
         model_name_lower = model_name.lower()
         
-        # Get today's failures
-        today_row = today_data[today_data.index == model_name_lower]
-        if today_row.empty:
-            continue
-        today_row = today_row.iloc[0]
+        # Get CURRENT failures from current_df
+        current_row = current_df.loc[model_name]
         
-        # Get yesterday's failures (if exists)
+        # Get YESTERDAY's failures from historical_df
         yesterday_row = yesterday_data[yesterday_data.index == model_name_lower]
         yesterday_failures_amd = {}
         yesterday_failures_nvidia = {}
@@ -491,33 +482,33 @@ def find_new_regressions(current_df: pd.DataFrame, historical_df: pd.DataFrame)
                 except:
                     yesterday_failures_nvidia = {}
         
-        # Get today's failures
-        today_failures_amd = today_row.get('failures_amd', {})
-        today_failures_nvidia = today_row.get('failures_nvidia', {})
+        # Get CURRENT failures
+        current_failures_amd = current_row.get('failures_amd', {})
+        current_failures_nvidia = current_row.get('failures_nvidia', {})
         
         # Handle string/dict conversion
-        if isinstance(today_failures_amd, str):
+        if isinstance(current_failures_amd, str):
             try:
-                today_failures_amd = json.loads(today_failures_amd)
+                current_failures_amd = json.loads(current_failures_amd)
             except:
-                today_failures_amd = {}
-        if isinstance(today_failures_nvidia, str):
+                current_failures_amd = {}
+        if isinstance(current_failures_nvidia, str):
             try:
-                today_failures_nvidia = json.loads(today_failures_nvidia)
+                current_failures_nvidia = json.loads(current_failures_nvidia)
             except:
-                today_failures_nvidia = {}
+                current_failures_nvidia = {}
         
-        # Check AMD failures
+        # Check AMD failures - find tests failing NOW but NOT yesterday
         for gpu_type in ['single', 'multi']:
-            today_tests = today_failures_amd.get(gpu_type, [])
+            current_tests = current_failures_amd.get(gpu_type, [])
             yesterday_tests = yesterday_failures_amd.get(gpu_type, [])
             
             # Get test names
-            today_test_names = {test.get('line', '') for test in today_tests}
+            current_test_names = {test.get('line', '') for test in current_tests}
             yesterday_test_names = {test.get('line', '') for test in yesterday_tests}
             
-            # Find new failures
-            new_tests = today_test_names - yesterday_test_names
+            # Find NEW failures: failing NOW but NOT yesterday
+            new_tests = current_test_names - yesterday_test_names
             for test_name in new_tests:
                 if test_name:  # Skip empty names
                     new_regressions.append({
@@ -528,17 +519,17 @@ def find_new_regressions(current_df: pd.DataFrame, historical_df: pd.DataFrame)
                         'gpu_type': gpu_type
                     })
         
-        # Check NVIDIA failures
+        # Check NVIDIA failures - find tests failing NOW but NOT yesterday
         for gpu_type in ['single', 'multi']:
-            today_tests = today_failures_nvidia.get(gpu_type, [])
+            current_tests = current_failures_nvidia.get(gpu_type, [])
             yesterday_tests = yesterday_failures_nvidia.get(gpu_type, [])
             
             # Get test names
-            today_test_names = {test.get('line', '') for test in today_tests}
+            current_test_names = {test.get('line', '') for test in current_tests}
             yesterday_test_names = {test.get('line', '') for test in yesterday_tests}
             
-            # Find new failures
-            new_tests = today_test_names - yesterday_test_names
+            # Find NEW failures: failing NOW but NOT yesterday
+            new_tests = current_test_names - yesterday_test_names
             for test_name in new_tests:
                 if test_name:  # Skip empty names
                     new_regressions.append({