Fix error when parsing usage details for video without audio track in Google models (#2507)

Author: jerry-heygen

pydantic_ai_slim/pydantic_ai/models/gemini.py

@@ -878,7 +878,7 @@ def _metadata_as_usage(response: _GeminiResponse) -> usage.Usage:
             metadata_details = cast(list[_GeminiModalityTokenCount], metadata_details)
             suffix = key.removesuffix('_details')
             for detail in metadata_details:
-                details[f'{detail["modality"].lower()}_{suffix}'] = detail['token_count']
+                details[f'{detail["modality"].lower()}_{suffix}'] = detail.get('token_count', 0)
 
     return usage.Usage(
         request_tokens=metadata.get('prompt_token_count', 0),

pydantic_ai_slim/pydantic_ai/models/google.py

@@ -603,7 +603,7 @@ def _metadata_as_usage(response: GenerateContentResponse) -> usage.Usage:
         if key.endswith('_details') and metadata_details:
             suffix = key.removesuffix('_details')
             for detail in metadata_details:
-                details[f'{detail["modality"].lower()}_{suffix}'] = detail['token_count']
+                details[f'{detail["modality"].lower()}_{suffix}'] = detail.get('token_count', 0)
 
     return usage.Usage(
         request_tokens=metadata.get('prompt_token_count', 0),