refactor: remove explicit None assignments for usage token fields

packages/image-generation/src/celeste_image_generation/providers/bytedance/streaming.py

@@ -38,8 +38,6 @@ def _parse_chunk(self, chunk_data: dict[str, Any]) -> ImageGenerationChunk | Non
             if usage_data:
                 self._completed_usage = ImageGenerationUsage(
                     total_tokens=usage_data.get("total_tokens"),
-                    input_tokens=None,
-                    output_tokens=None,
                 )
             return None
 

packages/text-generation/src/celeste_text_generation/providers/anthropic/client.py

@@ -49,9 +49,7 @@ def _parse_usage(self, response_data: dict[str, Any]) -> TextGenerationUsage:
             input_tokens=input_tokens,
             output_tokens=output_tokens,
             total_tokens=total_tokens,
-            billed_tokens=None,
-            cached_tokens=None,
-            reasoning_tokens=None,
+            cached_tokens=usage_data.get("cache_read_input_tokens"),
         )
 
     def _parse_content(

packages/text-generation/src/celeste_text_generation/providers/anthropic/streaming.py

@@ -105,9 +105,7 @@ def _parse_usage_from_event(
             input_tokens=input_tokens,
             output_tokens=output_tokens,
             total_tokens=total_tokens,
-            billed_tokens=None,
-            cached_tokens=None,
-            reasoning_tokens=None,
+            cached_tokens=usage_data.get("cache_read_input_tokens"),
         )
 
     def _parse_usage(self, chunks: list[TextGenerationChunk]) -> TextGenerationUsage:

packages/text-generation/src/celeste_text_generation/providers/google/client.py

@@ -48,8 +48,6 @@ def _parse_usage(self, response_data: dict[str, Any]) -> TextGenerationUsage:
             output_tokens=usage_metadata.get("candidatesTokenCount"),
             total_tokens=usage_metadata.get("totalTokenCount"),
             reasoning_tokens=usage_metadata.get("thoughtsTokenCount"),
-            billed_tokens=None,
-            cached_tokens=None,
         )
 
     def _parse_content(

packages/text-generation/src/celeste_text_generation/providers/mistral/client.py

@@ -62,9 +62,17 @@ def _parse_content(
 
         first_choice = choices[0]
         message = first_choice.get("message", {})
-        content = message.get("content") or ""
+        content = message.get("content")
 
-        return self._transform_output(content, **parameters)
+        # Handle magistral thinking models that return list content
+        if isinstance(content, list):
+            text_parts = []
+            for block in content:
+                if isinstance(block, dict) and block.get("type") == "text":
+                    text_parts.append(block.get("text", ""))
+            content = "".join(text_parts)
+
+        return self._transform_output(content or "", **parameters)
 
     def _parse_finish_reason(
         self, response_data: dict[str, Any]

packages/text-generation/src/celeste_text_generation/providers/mistral/streaming.py

@@ -56,6 +56,15 @@ def _parse_chunk(self, event: dict[str, Any]) -> TextGenerationChunk | None:
 
         # Extract content delta
         content_delta = delta.get("content")
+
+        # Handle magistral thinking models that may return list content
+        if isinstance(content_delta, list):
+            text_parts = []
+            for block in content_delta:
+                if isinstance(block, dict) and block.get("type") == "text":
+                    text_parts.append(block.get("text", ""))
+            content_delta = "".join(text_parts) if text_parts else None
+
         finish_reason_str = first_choice.get("finish_reason")
 
         # Extract usage from event if present (in final event)

packages/text-generation/src/celeste_text_generation/providers/openai/client.py

@@ -44,7 +44,6 @@ def _parse_usage(self, response_data: dict[str, Any]) -> TextGenerationUsage:
             total_tokens=usage_data.get("total_tokens"),
             cached_tokens=input_tokens_details.get("cached_tokens"),
             reasoning_tokens=output_tokens_details.get("reasoning_tokens"),
-            billed_tokens=None,
         )
 
     def _parse_content(

packages/text-generation/src/celeste_text_generation/providers/openai/streaming.py

@@ -59,7 +59,6 @@ def _parse_chunk(self, event: dict[str, Any]) -> TextGenerationChunk | None:
                     total_tokens=usage_data.get("total_tokens"),
                     cached_tokens=input_tokens_details.get("cached_tokens"),
                     reasoning_tokens=output_tokens_details.get("reasoning_tokens"),
-                    billed_tokens=None,
                 )
 
             finish_reason: TextGenerationFinishReason | None = None

packages/text-generation/src/celeste_text_generation/providers/xai/client.py

@@ -51,7 +51,6 @@ def _parse_usage(self, response_data: dict[str, Any]) -> TextGenerationUsage:
             total_tokens=usage_data.get("total_tokens"),
             cached_tokens=prompt_tokens_details.get("cached_tokens"),
             reasoning_tokens=completion_tokens_details.get("reasoning_tokens"),
-            billed_tokens=None,
         )
 
     def _parse_content(

packages/text-generation/src/celeste_text_generation/providers/xai/streaming.py

@@ -68,7 +68,6 @@ def _parse_chunk(self, event: dict[str, Any]) -> TextGenerationChunk | None:
                 total_tokens=usage_dict.get("total_tokens"),
                 cached_tokens=prompt_tokens_details.get("cached_tokens"),
                 reasoning_tokens=completion_tokens_details.get("reasoning_tokens"),
-                billed_tokens=None,
             )
 
         # Create finish reason if present