Merge pull request #23500 from BerriAI/litellm_litellm-mypy-errors-28de

[Fix] MyPy Errors

litellm/__init__.py

@@ -1897,7 +1897,7 @@ if TYPE_CHECKING:
     supports_reasoning: Callable[..., bool]
     acreate: Callable[..., Any]
     get_max_tokens: Callable[..., int]
-    get_model_info: Callable[..., _ModelInfoType]
+    get_model_info: Callable[..., _ModelInfoType]  # type: ignore[no-redef]
     register_prompt_template: Callable[..., None]
     validate_environment: Callable[..., dict]
     check_valid_key: Callable[..., bool]

litellm/litellm_core_utils/prompt_templates/factory.py

@@ -2283,7 +2283,7 @@ def sanitize_messages_for_tool_calling(
     for idx, msg in enumerate(sanitized_messages):
         role = msg.get("role")
         tcid = msg.get("tool_call_id") if role in ["tool", "function"] else None
-        if tcid:
+        if tcid and isinstance(tcid, str):
             if tcid in seen_in_block:
                 # Mark the earlier occurrence for removal (keep latest)
                 duplicates_to_remove.add(seen_in_block[tcid])
@@ -2581,13 +2581,11 @@ def anthropic_messages_pt(  # noqa: PLR0915
 
                 # Build the text block if content is a non-empty string
                 text_element = None
-                if (
-                    isinstance(assistant_content_block.get("content"), str)
-                    and assistant_content_block["content"]
-                ):
+                _acb_content = assistant_content_block.get("content")
+                if isinstance(_acb_content, str) and _acb_content:
                     _anthropic_text_content_element = AnthropicMessagesTextParam(
                         type="text",
-                        text=assistant_content_block["content"],
+                        text=_acb_content,
                     )
                     _content_element = add_cache_control_to_content(
                         anthropic_content_element=_anthropic_text_content_element,
@@ -2682,9 +2680,10 @@ def anthropic_messages_pt(  # noqa: PLR0915
                 _content_is_list = "content" in assistant_content_block and isinstance(
                     assistant_content_block["content"], list
                 )
+                _content_list = assistant_content_block.get("content") if _content_is_list else None
                 _list_has_thinking = False
-                if _content_is_list:
-                    for _item in assistant_content_block["content"]:
+                if _content_is_list and _content_list is not None:
+                    for _item in _content_list:
                         if isinstance(_item, dict) and _item.get("type") in (
                             "thinking",
                             "redacted_thinking",
@@ -2696,8 +2695,10 @@ def anthropic_messages_pt(  # noqa: PLR0915
                     thinking_blocks is not None and not _list_has_thinking
                 ):  # IMPORTANT: ADD THIS FIRST, ELSE ANTHROPIC WILL RAISE AN ERROR
                     assistant_content.extend(thinking_blocks)
-                if _content_is_list:
-                    for m in assistant_content_block["content"]:
+                if _content_is_list and _content_list is not None:
+                    for m in _content_list:
+                        if not isinstance(m, dict):
+                            continue
                         # handle thinking blocks
                         thinking_block = cast(str, m.get("thinking", ""))
                         text_block = cast(str, m.get("text", ""))

litellm/llms/anthropic/files/transformation.py

@@ -14,7 +14,7 @@ Anthropic Files API endpoints:
 
 import calendar
 import time
-from typing import Any, Dict, List, Optional, Union
+from typing import Any, Dict, List, Optional, Union, cast
 
 import httpx
 from openai.types.file_deleted import FileDeleted
@@ -79,7 +79,7 @@ class AnthropicFilesConfig(BaseFilesConfig):
         return AnthropicError(
             status_code=status_code,
             message=error_message,
-            headers=headers,
+            headers=cast(httpx.Headers, headers) if isinstance(headers, dict) else headers,
         )
 
     def validate_environment(

litellm/llms/black_forest_labs/image_edit/transformation.py

@@ -230,8 +230,8 @@ class BlackForestLabsImageEditConfig(BaseImageEditConfig):
     def transform_image_edit_request(
         self,
         model: str,
-        prompt: str,
-        image: FileTypes,
+        prompt: Optional[str],
+        image: Optional[FileTypes],
         image_edit_optional_request_params: Dict,
         litellm_params: GenericLiteLLMParams,
         headers: dict,

litellm/llms/black_forest_labs/image_generation/transformation.py

@@ -259,7 +259,12 @@ class BlackForestLabsImageGenerationConfig(BaseImageGenerationConfig):
         raw_response: httpx.Response,
         model_response: ImageResponse,
         logging_obj: LiteLLMLoggingObj,
-        **kwargs,
+        request_data: dict,
+        optional_params: dict,
+        litellm_params: dict,
+        encoding: Any,
+        api_key: Optional[str] = None,
+        json_mode: Optional[bool] = None,
     ) -> ImageResponse:
         """
         Transform Black Forest Labs response to OpenAI-compatible ImageResponse.

litellm/llms/brave/search/transformation.py

@@ -5,7 +5,7 @@ Documentation: https://api-dashboard.search.brave.com/app/documentation/web-sear
 
 from __future__ import annotations
 from datetime import datetime, timezone
-from dateutil import parser
+from dateutil import parser  # type: ignore[import-untyped]
 from typing import Dict, List, Literal, Optional, TypedDict, Union
 import httpx
 import re

litellm/llms/custom_httpx/llm_http_handler.py

@@ -3036,10 +3036,11 @@ class BaseLLMHTTPHandler:
         elif isinstance(transformed_request, dict) and "file" in transformed_request:
             # Handle multipart form-data uploads (e.g., Anthropic Files API)
             # The dict contains tuples suitable for httpx's `files` parameter
+            file_request = cast(Dict[str, Any], transformed_request)
             upload_response = sync_httpx_client.post(
                 url=api_base,
                 headers=headers,
-                files=transformed_request,
+                files=file_request,
                 timeout=timeout,
             )
         else:
@@ -4870,10 +4871,12 @@ class BaseLLMHTTPHandler:
                 timeout=timeout,
             )
         except Exception as e:
-            raise self._handle_error(
-                e=e,
-                provider_config=provider_config,
-            )
+            if provider_config is not None:
+                raise self._handle_error(
+                    e=e,
+                    provider_config=provider_config,
+                )
+            raise
 
     async def async_realtime_calls_handler(
         self,
@@ -4954,10 +4957,12 @@ class BaseLLMHTTPHandler:
                 timeout=timeout,
             )
         except Exception as e:
-            raise self._handle_error(
-                e=e,
-                provider_config=provider_config,
-            )
+            if provider_config is not None:
+                raise self._handle_error(
+                    e=e,
+                    provider_config=provider_config,
+                )
+            raise
 
     async def async_responses_websocket(
         self,
@@ -8026,7 +8031,7 @@ class BaseLLMHTTPHandler:
 
         url = api_base
 
-        params = {}
+        params: Dict[str, Any] = {}
         if after is not None:
             params["after"] = after
         if before is not None:
@@ -8108,7 +8113,7 @@ class BaseLLMHTTPHandler:
 
         url = api_base
 
-        params = {}
+        params: Dict[str, Any] = {}
         if after is not None:
             params["after"] = after
         if before is not None:
@@ -8170,7 +8175,7 @@ class BaseLLMHTTPHandler:
 
         url = f"{api_base}/{vector_store_id}"
 
-        request_body = dict(vector_store_update_optional_params)
+        request_body: Dict[str, Any] = dict(vector_store_update_optional_params)
 
         # Clean metadata to only include string values (OpenAI requirement)
         if "metadata" in request_body and request_body["metadata"] is not None:
@@ -8253,7 +8258,7 @@ class BaseLLMHTTPHandler:
 
         url = f"{api_base}/{vector_store_id}"
 
-        request_body = dict(vector_store_update_optional_params)
+        request_body: Dict[str, Any] = dict(vector_store_update_optional_params)
 
         # Clean metadata to only include string values (OpenAI requirement)
         if "metadata" in request_body and request_body["metadata"] is not None:

litellm/llms/perplexity/responses/transformation.py

@@ -63,16 +63,19 @@ class PerplexityResponsesConfig(OpenAIResponsesAPIConfig):
 
     def _ensure_message_type(
         self, input: Union[str, ResponseInputParam]
-    ) -> Union[str, List[Dict[str, Any]]]:
+    ) -> Union[str, ResponseInputParam]:
         """Ensure list input items have type='message' (required by Perplexity)."""
         if isinstance(input, str):
             return input
         if isinstance(input, list):
-            result = []
+            result: List[Any] = []
             for item in input:
                 if isinstance(item, dict) and "type" not in item:
-                    item = {**item, "type": "message"}
-                result.append(item)
+                    new_item = dict(item)  # convert to plain dict to avoid TypedDict checking
+                    new_item["type"] = "message"
+                    result.append(new_item)
+                else:
+                    result.append(item)
             return result
         return input
 

litellm/llms/vertex_ai/gemini_embeddings/batch_embed_content_handler.py

@@ -153,6 +153,7 @@ class GoogleBatchEmbeddings(VertexLLM):
 
         is_multimodal = _is_multimodal_input(input)
         use_embed_content = is_multimodal or (custom_llm_provider == "vertex_ai")
+        mode: Literal["embedding", "batch_embedding"]
         if use_embed_content:
             mode = "embedding"
         else:
@@ -200,6 +201,7 @@ class GoogleBatchEmbeddings(VertexLLM):
             )
 
         ### TRANSFORMATION (sync path) ###
+        request_data: Any
         if use_embed_content:
             resolved_files = {}
             if api_key:

litellm/llms/vertex_ai/vertex_ai_partner_models/llama3/transformation.py

@@ -190,7 +190,7 @@ class VertexAILlama3StreamingHandler(OpenAIChatCompletionStreamingHandler):
                     ],
                 )
                 # Modify current chunk to be the first chunk with role but no finish_reason
-                result.choices[0].finish_reason = None
+                result.choices[0].finish_reason = None  # type: ignore[assignment]
                 delta.role = "assistant"
                 # Ensure content is empty string for first chunk, not None
                 if delta.content is None:

litellm/main.py

@@ -7700,7 +7700,7 @@ async def acount_tokens(
     local_count = litellm.token_counter(
         model=model,
         messages=fallback_messages,
-        tools=tools,
+        tools=tools,  # type: ignore[arg-type]
     )
 
     return TokenCountResponse(

litellm/proxy/_experimental/mcp_server/db.py

@@ -142,9 +142,9 @@ def decrypt_credentials(
         "aws_session_token",
     ]
     for field in secret_fields:
-        value = credentials.get(field)
-        if value is not None:
-            credentials[field] = decrypt_value_helper(
+        value = credentials.get(field)  # type: ignore[literal-required]
+        if value is not None and isinstance(value, str):
+            credentials[field] = decrypt_value_helper(  # type: ignore[literal-required]
                 value=value,
                 key=field,
                 exception_type="debug",

litellm/proxy/common_request_processing.py

@@ -530,6 +530,8 @@ class ProxyBaseLLMRequestProcessing:
             "aresponses",
             "_arealtime",
             "_aresponses_websocket",
+            "acreate_realtime_client_secret",
+            "arealtime_calls",
             "aget_responses",
             "adelete_responses",
             "acancel_responses",
@@ -553,6 +555,10 @@ class ProxyBaseLLMRequestProcessing:
             "allm_passthrough_route",
             "avector_store_search",
             "avector_store_create",
+            "avector_store_retrieve",
+            "avector_store_list",
+            "avector_store_update",
+            "avector_store_delete",
             "avector_store_file_create",
             "avector_store_file_list",
             "avector_store_file_retrieve",
@@ -774,20 +780,36 @@ class ProxyBaseLLMRequestProcessing:
             "aembedding",
             "aresponses",
             "_arealtime",
+            "_aresponses_websocket",
             "acreate_realtime_client_secret",
             "arealtime_calls",
             "aget_responses",
             "adelete_responses",
             "acancel_responses",
             "acompact_responses",
+            "acreate_batch",
+            "aretrieve_batch",
+            "alist_batches",
+            "acancel_batch",
+            "afile_content",
+            "afile_retrieve",
+            "afile_delete",
             "atext_completion",
-            "aimage_edit",
+            "acreate_fine_tuning_job",
+            "acancel_fine_tuning_job",
+            "alist_fine_tuning_jobs",
+            "aretrieve_fine_tuning_job",
             "alist_input_items",
+            "aimage_edit",
             "agenerate_content",
             "agenerate_content_stream",
             "allm_passthrough_route",
             "avector_store_search",
             "avector_store_create",
+            "avector_store_retrieve",
+            "avector_store_list",
+            "avector_store_update",
+            "avector_store_delete",
             "avector_store_file_create",
             "avector_store_file_list",
             "avector_store_file_retrieve",
@@ -815,8 +837,8 @@ class ProxyBaseLLMRequestProcessing:
             "aget_interaction",
             "adelete_interaction",
             "acancel_interaction",
-            "acancel_batch",
-            "afile_delete",
+            "asend_message",
+            "call_mcp_tool",
             "acreate_eval",
             "alist_evals",
             "aget_eval",

litellm/proxy/guardrails/guardrail_hooks/panw_prisma_airs/panw_prisma_airs.py

@@ -315,6 +315,7 @@ class PanwPrismaAirsHandler(CustomGuardrail):
             panw_metadata["litellm_trace_id"] = metadata["litellm_trace_id"]
 
         # Build contents: tool_event takes priority, else prompt/response text
+        contents: List[Dict[str, Any]]
         if tool_event is not None:
             contents = [{"tool_event": tool_event}]
         else:
@@ -1485,7 +1486,7 @@ class PanwPrismaAirsHandler(CustomGuardrail):
             detail = (
                 e.detail if isinstance(e.detail, dict) else {"message": str(e.detail)}
             )
-            error_obj = dict(detail.get("error", detail))
+            error_obj: Dict[str, Any] = dict(detail.get("error", detail))  # type: ignore[arg-type]
             error_obj["code"] = e.status_code
             yield f"data: {json.dumps({'error': error_obj})}\n\n"
         except Exception as e:

litellm/proxy/guardrails/guardrail_hooks/presidio.py

@@ -106,9 +106,9 @@ class _OPTIONAL_PresidioPIIMasking(CustomGuardrail):
         if (self.output_parse_pii or self.apply_to_output) and not logging_only:
             current_hook = self.event_hook
             if isinstance(current_hook, str) and current_hook != "post_call":
-                self.event_hook = [current_hook, "post_call"]
+                self.event_hook = cast(List[GuardrailEventHooks], [current_hook, "post_call"])
             elif isinstance(current_hook, list) and "post_call" not in current_hook:
-                self.event_hook = current_hook + ["post_call"]
+                self.event_hook = cast(List[GuardrailEventHooks], current_hook + ["post_call"])
         self.pii_entities_config: Dict[Union[PiiEntityType, str], PiiAction] = (
             pii_entities_config or {}
         )
@@ -908,7 +908,7 @@ class _OPTIONAL_PresidioPIIMasking(CustomGuardrail):
         if self.apply_to_output is True:
             if self._is_anthropic_message_response(response):
                 return await self._process_anthropic_response_for_pii(
-                    response=response, request_data=data, mode="mask"
+                    response=cast(dict, response), request_data=data, mode="mask"
                 )
             return await self._mask_output_response(
                 response=response, request_data=data
@@ -927,7 +927,7 @@ class _OPTIONAL_PresidioPIIMasking(CustomGuardrail):
             )
         elif self._is_anthropic_message_response(response):
             await self._process_anthropic_response_for_pii(
-                response=response, request_data=data, mode="unmask"
+                response=cast(dict, response), request_data=data, mode="unmask"
             )
         return response
 

litellm/proxy/hooks/mcp_semantic_filter/hook.py

@@ -259,6 +259,7 @@ class SemanticToolFilterHook(CustomLogger):
         user_api_key_dict: "UserAPIKeyAuth",
         response: Any,
         request_headers: Optional[Dict[str, str]] = None,
+        litellm_call_info: Optional[Dict[str, Any]] = None,
     ) -> Optional[Dict[str, str]]:
         """Add semantic filter stats and tool names to response headers."""
         from litellm.constants import MAX_MCP_SEMANTIC_FILTER_TOOLS_HEADER_LENGTH

litellm/proxy/management_endpoints/key_management_endpoints.py

@@ -1776,11 +1776,13 @@ async def _validate_mcp_servers_for_key_update(
             user_api_key_cache=user_api_key_cache,
             check_db_only=True,
         )
-    object_permission_dict = (
-        data.object_permission.model_dump()
-        if hasattr(data.object_permission, "model_dump")
-        else data.object_permission
-    )
+    object_permission_dict: Optional[dict] = None
+    if data.object_permission is not None:
+        object_permission_dict = (
+            data.object_permission.model_dump()
+            if hasattr(data.object_permission, "model_dump")
+            else dict(data.object_permission)  # type: ignore[arg-type]
+        )
     await validate_key_mcp_servers_against_team(
         object_permission=object_permission_dict,
         team_obj=effective_team_obj,

litellm/proxy/route_llm_request.py

@@ -173,8 +173,21 @@ async def route_request(  # noqa: PLR0915 - Complex routing function, refactorin
         "agenerate_content",
         "agenerate_content_stream",
         "allm_passthrough_route",
+        "acreate_batch",
+        "aretrieve_batch",
+        "alist_batches",
+        "afile_content",
+        "afile_retrieve",
+        "acreate_fine_tuning_job",
+        "acancel_fine_tuning_job",
+        "alist_fine_tuning_jobs",
+        "aretrieve_fine_tuning_job",
         "avector_store_search",
         "avector_store_create",
+        "avector_store_retrieve",
+        "avector_store_list",
+        "avector_store_update",
+        "avector_store_delete",
         "avector_store_file_create",
         "avector_store_file_list",
         "avector_store_file_retrieve",
@@ -207,6 +220,8 @@ async def route_request(  # noqa: PLR0915 - Complex routing function, refactorin
         "aget_interaction",
         "adelete_interaction",
         "acancel_interaction",
+        "asend_message",
+        "call_mcp_tool",
         "acancel_batch",
         "afile_delete",
         "acreate_eval",

litellm/proxy/vector_store_endpoints/endpoints.py

@@ -386,7 +386,7 @@ async def vector_store_list(
         version,
     )
 
-    data = {}
+    data: dict = {}
     if after is not None:
         data["after"] = after
     if before is not None:

litellm/responses/litellm_completion_transformation/transformation.py

@@ -410,11 +410,12 @@ class LiteLLMCompletionResponsesConfig:
                                     else getattr(new_msg, "role", None)
                                 )
                                 if new_role == "assistant":
-                                    new_tcs = (
+                                    _raw_tcs = (
                                         new_msg.get("tool_calls")
                                         if isinstance(new_msg, dict)
                                         else getattr(new_msg, "tool_calls", None)
-                                    ) or []
+                                    )
+                                    new_tcs: list = _raw_tcs if isinstance(_raw_tcs, list) else []
                                     for tc in new_tcs:
                                         LiteLLMCompletionResponsesConfig._add_tool_call_to_assistant(
                                             last_msg, tc

litellm/types/google_genai/main.py

@@ -19,11 +19,11 @@ if TYPE_CHECKING:
     GenerateContentRequestParametersDict = _genai_types._GenerateContentParametersDict
     ToolConfigDict = _genai_types.ToolConfigDict
 
-    class GenerateContentRequestDict(GenerateContentRequestParametersDict):  # type: ignore[misc]
+    class GenerateContentRequestDict(GenerateContentRequestParametersDict):  # type: ignore[misc, valid-type]
         generationConfig: Optional[Any]
-        tools: Optional[ToolConfigDict]  # type: ignore[assignment]
+        tools: Optional[ToolConfigDict]  # type: ignore[assignment, valid-type]
 
-    class GenerateContentResponse(GoogleGenAIGenerateContentResponse, BaseLiteLLMOpenAIResponseObject):  # type: ignore[misc]
+    class GenerateContentResponse(GoogleGenAIGenerateContentResponse, BaseLiteLLMOpenAIResponseObject):  # type: ignore[misc, valid-type]
         _hidden_params: dict = {}
         pass
 

litellm/types/llms/openai.py

@@ -1052,6 +1052,14 @@ OpenAIImageGenerationOptionalParams = Literal[
     "size",
     "style",
     "user",
+    "seed",
+    "safety_tolerance",
+    "prompt_upsampling",
+    "raw",
+    "num_images",
+    "image_url",
+    "image_prompt_strength",
+    "aspect_ratio",
 ]
 
 OpenAIImageEditOptionalParams = Literal[

litellm/types/utils.py

@@ -1663,7 +1663,7 @@ class StreamingChoices(OpenAIObject):
         if finish_reason:
             self.finish_reason = map_finish_reason(finish_reason)
         else:
-            self.finish_reason = None
+            self.finish_reason = None  # type: ignore[assignment]
         self.index = index
         if delta is not None:
             if isinstance(delta, Delta):

litellm/utils.py

@@ -8141,6 +8141,8 @@ class ProviderConfigManager:
                 raise ValueError(f"Provider {provider.value} not found")
             return create_config_class(provider_config)()
 
+        return None
+
     @staticmethod
     def get_provider_embedding_config(
         model: str,

tests/test_litellm/llms/black_forest_labs/image_generation/test_bfl_image_generation_transformation.py

@@ -282,6 +282,10 @@ class TestBlackForestLabsImageGenerationTransformation:
             raw_response=mock_response,
             model_response=model_response,
             logging_obj=self.logging_obj,
+            request_data={},
+            optional_params={},
+            litellm_params={},
+            encoding=None,
         )
 
         assert len(result.data) == 1
@@ -306,6 +310,10 @@ class TestBlackForestLabsImageGenerationTransformation:
             raw_response=mock_response,
             model_response=model_response,
             logging_obj=self.logging_obj,
+            request_data={},
+            optional_params={},
+            litellm_params={},
+            encoding=None,
         )
 
         assert len(result.data) == 2
@@ -329,6 +337,10 @@ class TestBlackForestLabsImageGenerationTransformation:
                 raw_response=mock_response,
                 model_response=model_response,
                 logging_obj=self.logging_obj,
+                request_data={},
+                optional_params={},
+                litellm_params={},
+                encoding=None,
             )
 
     def test_get_error_class(self):