Fix failing tests

litellm/completion_extras/litellm_responses_transformation/transformation.py

@@ -625,7 +625,10 @@ class LiteLLMResponsesTransformationHandler(CompletionTransformationBridge):
         if not isinstance(item, dict):
             return
         try:
-            output_index = int(parsed_chunk.get("output_index"))
+            output_index_raw = parsed_chunk.get("output_index")
+            if output_index_raw is None:
+                raise ValueError("missing output_index")
+            output_index = int(output_index_raw)
         except (TypeError, ValueError):
             output_index = len(recovered_output_items)
         recovered_output_items[output_index] = item
@@ -642,7 +645,10 @@ class LiteLLMResponsesTransformationHandler(CompletionTransformationBridge):
             return
 
         try:
-            output_index = int(parsed_chunk.get("output_index"))
+            output_index_raw = parsed_chunk.get("output_index")
+            if output_index_raw is None:
+                raise ValueError("missing output_index")
+            output_index = int(output_index_raw)
         except (TypeError, ValueError):
             output_index = len(recovered_text_only_items)
 
@@ -664,7 +670,10 @@ class LiteLLMResponsesTransformationHandler(CompletionTransformationBridge):
             return
 
         try:
-            content_index = int(parsed_chunk.get("content_index"))
+            content_index_raw = parsed_chunk.get("content_index")
+            if content_index_raw is None:
+                raise ValueError("missing content_index")
+            content_index = int(content_index_raw)
         except (TypeError, ValueError):
             content_index = len(content)
 
@@ -771,8 +780,8 @@ class LiteLLMResponsesTransformationHandler(CompletionTransformationBridge):
                 logging_obj
             )
             if recovered_output_items:
-                output_items = recovered_output_items
-                raw_response.output = recovered_output_items
+                output_items = cast(Any, recovered_output_items)
+                raw_response.output = cast(Any, recovered_output_items)
                 verbose_logger.warning(
                     "Recovered empty Responses API output from raw SSE for model=%s",
                     model,

litellm/llms/chatgpt/responses/transformation.py

@@ -213,6 +213,8 @@ class ChatGPTResponsesAPIConfig(OpenAIResponsesAPIConfig):
         if not isinstance(item, dict):
             return
         try:
+            if output_index is None:
+                raise ValueError("missing output_index")
             index = int(output_index)
         except (TypeError, ValueError):
             index = len(streamed_output_items)

litellm/llms/reducto/common.py

@@ -131,11 +131,12 @@ def build_pages_from_reducto(result: Dict[str, Any]) -> List["OCRPage"]:
             block.get("content", "") for block in blocks if block.get("content")
         )
         page_index = max(page_no - 1, 0)
-        pages.append(
-            OCRPage(
-                index=page_index,
-                markdown=markdown,
-                blocks=blocks,
-            )
+        page = OCRPage(
+            index=page_index,
+            markdown=markdown,
         )
+        # OCRPage accepts extra keys at runtime; assign blocks after construction
+        # so static typing does not reject provider-specific metadata.
+        setattr(page, "blocks", blocks)
+        pages.append(page)
     return pages

litellm/llms/vertex_ai/vertex_llm_base.py

@@ -45,7 +45,7 @@ class VertexBase:
         self._credentials: Optional[GoogleCredentialsObject] = None
         self._credentials_project_mapping: Dict[
             Tuple[Optional[VERTEX_CREDENTIALS_TYPES], Optional[str]],
-            Tuple[GoogleCredentialsObject, str],
+            Tuple[GoogleCredentialsObject, Optional[str]],
         ] = {}
         self.project_id: Optional[str] = None
         self.async_handler: Optional[AsyncHTTPHandler] = None