refactor: lazy load litellm_core_utils & change bad import patterm

litellm/__init__.py

@@ -1053,8 +1053,6 @@ openai_image_generation_models = ["dall-e-2", "dall-e-3"]
 openai_video_generation_models = ["sora-2"]
 
 # timeout is lazy-loaded via __getattr__
-from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
-from litellm.litellm_core_utils.core_helpers import remove_index_from_tool_calls
 
 # Import KeyManagementSettings here (before utils import) because _key_management_settings
 # is accessed during import time in secret_managers/main.py (via dd_tracing -> datadog -> _service_logger -> utils)
@@ -1503,6 +1501,8 @@ if TYPE_CHECKING:
     get_first_chars_messages: Callable[..., str]
     get_provider_fields: Callable[..., List]
     get_valid_models: Callable[..., list]
+    get_llm_provider: Callable[..., Tuple[str, str, Optional[str], Optional[str]]]
+    remove_index_from_tool_calls: Callable[..., None]
 
     # Response types - truly lazy loaded only (not in main.py or elsewhere)
     ModelResponseListIterator: Type[Any]

litellm/_lazy_imports_registry.py

@@ -32,6 +32,7 @@ UTILS_NAMES = (
     "ModelResponse", "ModelResponseStream", "EmbeddingResponse", "ImageResponse",
     "TranscriptionResponse", "TextCompletionResponse", "get_provider_fields",
     "ModelResponseListIterator", "get_valid_models", "timeout",
+    "get_llm_provider", "remove_index_from_tool_calls",
 )
 
 # Token counter names that support lazy loading via _lazy_import_token_counter
@@ -330,6 +331,8 @@ _UTILS_IMPORT_MAP = {
     "ModelResponseListIterator": (".utils", "ModelResponseListIterator"),
     "get_valid_models": (".utils", "get_valid_models"),
     "timeout": (".timeout", "timeout"),
+    "get_llm_provider": ("litellm.litellm_core_utils.get_llm_provider_logic", "get_llm_provider"),
+    "remove_index_from_tool_calls": ("litellm.litellm_core_utils.core_helpers", "remove_index_from_tool_calls"),
 }
 
 _COST_CALCULATOR_IMPORT_MAP = {

litellm/realtime_api/main.py

@@ -3,7 +3,7 @@
 from typing import Any, Optional, cast
 
 import litellm
-from litellm import get_llm_provider
+from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
 from litellm.constants import REALTIME_WEBSOCKET_MAX_MESSAGE_SIZE_BYTES
 from litellm.llms.base_llm.realtime.transformation import BaseRealtimeConfig
 from litellm.llms.custom_httpx.llm_http_handler import BaseLLMHTTPHandler

litellm/router_utils/pattern_match_deployments.py

@@ -7,7 +7,7 @@ import re
 from re import Match
 from typing import Dict, List, Optional, Tuple
 
-from litellm import get_llm_provider
+from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
 from litellm._logging import verbose_router_logger