Merge f736ac8891 into 17db64ef78

litellm/anthropic_beta_headers_manager.py

@@ -189,7 +189,6 @@ def _load_beta_headers_config() -> dict:
         if _BETA_HEADERS_CONFIG is not None:
             return _BETA_HEADERS_CONFIG
 
-        # Get the URL from environment or use default
         from litellm import anthropic_beta_headers_url
 
         _BETA_HEADERS_CONFIG = get_beta_headers_config(url=anthropic_beta_headers_url)

litellm/litellm_core_utils/get_model_cost_map.py

@@ -318,9 +318,7 @@ def _classify_fetch_content(content: bytes, url: str, etag: str | None) -> _Fetc
         return ModelCostMapReloadUnavailable(reason=f"invalid JSON from {url}: {e}")
     if not isinstance(parsed, dict):
         return ModelCostMapReloadUnavailable(reason=f"expected a JSON object from {url}, got {type(parsed).__name__}")
-    model_cost_map: Final = cast(  # cast-ok: JSON object keys are strings by definition
-        dict[str, object], parsed
-    )
+    model_cost_map: Final = _BUNDLED_CATALOG_ADAPTER.validate_python(parsed)
     return ModelCostMapReloaded(model_cost_map=model_cost_map, revision=git_blob_id(content), etag=etag)
 
 

litellm/litellm_core_utils/litellm_logging.py

@@ -2757,7 +2757,7 @@ class Logging(LiteLLMLoggingBaseClass):
         response: ModelResponse | TextCompletionResponse | ResponsesAPIResponse | InteractionsAPIResponse,
     ) -> ModelResponse | TextCompletionResponse | ResponsesAPIResponse | InteractionsAPIResponse:
         response_id: Final = self.model_call_details.get("continued_response_id")
-        if not isinstance(response_id, str) or getattr(response, "id", None) == response_id:
+        if not isinstance(response_id, str) or response.id == response_id:
             return response
         return response.model_copy(update={"id": response_id})
 
@@ -5739,7 +5739,7 @@ def _init_custom_logger_compatible_class(
     logging_integration: str,
     internal_usage_cache: DualCache | None,
     llm_router: object,
-    custom_logger_init_args: dict[str, object] | None = {},
+    custom_logger_init_args: dict[str, object] | None = None,
     raise_import_errors: bool = False,
 ) -> CustomLogger | None:
     """

litellm/litellm_core_utils/streaming_handler.py

@@ -2464,8 +2464,6 @@ class CustomStreamWrapper:
         return any(get(field) for field in cls._CONTINUATION_DISQUALIFYING_DELTA_FIELDS)
 
     def _accumulate_streamed_delta(self, delta: object) -> None:
-        # Shared by the sync, async, and non-aiohttp iteration sites so answer
-        # text and the disqualifying-content latch stay in step across all three.
         get: Final = getattr(delta, "get", None)
         content: Final = get("content") if callable(get) else None
         self.response_uptil_now += content if isinstance(content, str) else ""

litellm/llms/anthropic/pass_through/caller_model_checks.py

@@ -243,8 +243,7 @@ async def caller_within_model_budget(
         return "allowed"
 
     model_max_budget: Final = getattr(user_api_key_auth, "model_max_budget", None)
-    token: Final = getattr(user_api_key_auth, "token", None)
-    if isinstance(model_max_budget, dict) and model_max_budget and token is not None:
+    if isinstance(model_max_budget, dict) and model_max_budget and user_api_key_auth.token is not None:
         try:
             await model_max_budget_limiter.is_key_within_model_budget(
                 user_api_key_dict=user_api_key_auth,
@@ -299,10 +298,8 @@ async def caller_within_model_budget(
             )
             return "denied_for_now"
 
-    end_user_model_max_budget: Final[dict[str, object] | None] = getattr(
-        user_api_key_auth, "end_user_model_max_budget", None
-    )
-    end_user_id: Final[str | None] = getattr(user_api_key_auth, "end_user_id", None)
+    end_user_model_max_budget: Final[dict[str, object] | None] = user_api_key_auth.end_user_model_max_budget
+    end_user_id: Final[str | None] = user_api_key_auth.end_user_id
     if isinstance(end_user_model_max_budget, dict) and end_user_model_max_budget and end_user_id is not None:
         try:
             await model_max_budget_limiter.is_end_user_within_model_budget(

litellm/llms/bedrock/count_tokens/kimi_k3.py

@@ -452,9 +452,7 @@ def _converse_request(
     if not all(_text_only(message) for message in messages):
         return None
     try:
-        chat_tools: Final = tuple(
-            chat_tool for tool in tools or () for chat_tool in _as_converse_input(tool)
-        )  # comprehension-ok: flattens each tool's translation
+        chat_tools: Final = tuple(chain.from_iterable(_as_converse_input(tool) for tool in tools or ()))
         return _ConverseRequest.model_validate(
             litellm.AmazonConverseConfig().transform_request(
                 model=model,

litellm/main.py

@@ -5894,7 +5894,6 @@ def completion(
                 return resolve_timeout(stream_kind=StreamKind.NOT_STREAMED)
             return resolve_timeout(stream_kind=StreamKind.FAKED if fakes_stream(attempt_params) else StreamKind.LIVE)
 
-        ### TIMEOUT LOGIC ###
         timeout = timeout_for_optional_params(optional_params)
         logging.record_applied_timeout(timeout)
         processed_non_default_params: Final = pre_process_non_default_params(

litellm/proxy/_experimental/mcp_server/mcp_server_manager.py

@@ -6187,9 +6187,9 @@ class MCPServerManager:
             "server_name": server_name,
             "mcp_rate_limit_server_name": server.alias or server.server_name or server.name,
             "user_api_key_auth": checked_auth,
-            "user_api_key_user_id": (getattr(checked_auth, "user_id", None) if checked_auth else None),
-            "user_api_key_team_id": (getattr(checked_auth, "team_id", None) if checked_auth else None),
-            "user_api_key_end_user_id": (getattr(checked_auth, "end_user_id", None) if checked_auth else None),
+            "user_api_key_user_id": checked_auth.user_id if checked_auth else None,
+            "user_api_key_team_id": checked_auth.team_id if checked_auth else None,
+            "user_api_key_end_user_id": checked_auth.end_user_id if checked_auth else None,
             "user_api_key_hash": (getattr(checked_auth, "api_key_hash", None) if checked_auth else None),
             "incoming_bearer_token": caller_bearer.get_secret_value() if caller_bearer is not None else None,
             "headers": logging_safe_mcp_headers(raw_headers),

litellm/proxy/db/db_transaction_queue/spend_log_cleanup.py

@@ -595,7 +595,7 @@ class SpendLogCleanup:
                 rows: Final = _STORAGE_ROWS.validate_python(
                     await tx.query_raw(trusted_sql(_STORAGE_SQL), list(_SPEND_LOG_TABLES))
                 )
-        except Exception as e:  # noqa: BLE001 - a failed measurement skips the size pass, it must not fail the run
+        except Exception as e:  # noqa: BLE001  # a failed measurement skips the size pass, it must not fail the run
             verbose_proxy_logger.warning("Could not measure spend-log storage: %s: %s", type(e).__name__, e)
             return None
         if not rows:
@@ -663,7 +663,7 @@ class SpendLogCleanup:
                     await tx.query_raw(trusted_sql(_LOAD_SIZE_STATE_SQL), SPEND_LOG_SIZE_CLEANUP_STATE_PARAM)
                 )
             return SizeCleanupState.model_validate_json(rows[0].state_json) if rows else None
-        except Exception as e:  # noqa: BLE001 - an unreadable count restarts from the catalog, it must not fail the run
+        except Exception as e:  # noqa: BLE001  # an unreadable count restarts from the catalog, it must not fail the run
             verbose_proxy_logger.warning("Could not read the spend-log size cleanup state: %s: %s", type(e).__name__, e)
             return None
 
@@ -675,7 +675,7 @@ class SpendLogCleanup:
                 await tx.execute_raw(
                     trusted_sql(_SAVE_SIZE_STATE_SQL), SPEND_LOG_SIZE_CLEANUP_STATE_PARAM, state.model_dump_json()
                 )
-        except Exception as e:  # noqa: BLE001 - a lost save costs one re-seed from the catalog, it must not fail the run
+        except Exception as e:  # noqa: BLE001  # a lost save costs one re-seed from the catalog, it must not fail the run
             verbose_proxy_logger.warning("Could not save the spend-log size cleanup state: %s: %s", type(e).__name__, e)
 
     async def _size_step_cutoff(self, prisma_client: PrismaClient, deadline: float, rows: int) -> datetime | None:

litellm/proxy/guardrails/guardrail_hooks/generic_guardrail_api/__init__.py

@@ -32,7 +32,7 @@ def initialize_guardrail(litellm_params: "LitellmParams", guardrail: "Guardrail"
         additional_provider_specific_params=getattr(litellm_params, "additional_provider_specific_params", {}),
         unreachable_fallback=getattr(litellm_params, "unreachable_fallback", "fail_closed"),
         fail_on_error=getattr(litellm_params, "fail_on_error", True),
-        forward_messages_as_sent=getattr(litellm_params, "forward_messages_as_sent", False),
+        forward_messages_as_sent=litellm_params.forward_messages_as_sent,
         extra_headers=getattr(litellm_params, "extra_headers", None),
         guardrail_name=guardrail.get("guardrail_name", ""),
         event_hook=litellm_params.mode,

litellm/proxy/proxy_server.py

@@ -16161,9 +16161,6 @@ async def _apply_search_filter_to_models(
         team_public_model_name: Final = (m.get("model_info") or {}).get("team_public_model_name") or ""
         if search_lower in team_public_model_name.lower():
             return True
-        # Also match the underlying LiteLLM model name (e.g.
-        # "openrouter/deepseek/deepseek-chat"), so users can find
-        # deployments by typing the provider or the upstream model id.
         litellm_params: Final[Mapping[str, object]] = m.get("litellm_params") or MappingProxyType({})
         litellm_model: Final = litellm_params.get("model")
         return isinstance(litellm_model, str) and search_lower in litellm_model.lower()

litellm/proxy/utils.py

@@ -27,7 +27,7 @@ from dataclasses import dataclass, field
 from datetime import date, datetime, timedelta, timezone
 from email.mime.multipart import MIMEMultipart
 from email.mime.text import MIMEText
-from functools import partial
+from functools import partial, reduce
 from itertools import takewhile
 from operator import attrgetter
 from types import MappingProxyType
@@ -1396,14 +1396,17 @@ _LOG_DB_METRICS_CALL_TYPES: Final = frozenset(("get_data", "insert_data", "updat
 _EXCEPTION_CHAIN_MAX_DEPTH: Final = 16
 
 
+def _extended_exception_chain(chain: tuple[BaseException, ...], iteration: int) -> tuple[BaseException, ...]:
+    if len(chain) != iteration + 1:
+        return chain
+    following: Final = chain[-1].__cause__ or chain[-1].__context__
+    if following is None or following in chain:
+        return chain
+    return (*chain, following)
+
+
 def _exception_chain(exception: BaseException) -> tuple[BaseException, ...]:
-    chain: Final[list[BaseException]] = [exception]  # mutable-ok: built once, returned frozen
-    for _ in range(_EXCEPTION_CHAIN_MAX_DEPTH):
-        following = chain[-1].__cause__ or chain[-1].__context__
-        if following is None or following in chain:
-            break
-        chain.append(following)
-    return tuple(chain)
+    return reduce(_extended_exception_chain, range(_EXCEPTION_CHAIN_MAX_DEPTH), (exception,))
 
 
 def _timed_out_reading_from_provider(exception: Exception) -> bool: