fix(logging): stop the spend metadata merge from walking a dict the logging thread writes to (#46044)

litellm/litellm_core_utils/core_helpers.py

@@ -318,7 +318,7 @@ def remove_items_at_indices(items: _PoppableByIndex | None, indices: Iterable[in
             items.pop(index)
 
 
-def add_missing_spend_metadata_to_litellm_metadata(litellm_metadata: dict, metadata: dict) -> dict:
+def add_missing_spend_metadata_to_litellm_metadata(litellm_metadata: dict, metadata: Mapping[str, object]) -> dict:
     """
     Helper to get litellm metadata for spend tracking
 
@@ -326,7 +326,7 @@ def add_missing_spend_metadata_to_litellm_metadata(litellm_metadata: dict, metad
     and user_api_key values are in 'metadata'.
     """
     potential_spend_tracking_metadata_substring: Final = "user_api_key"
-    for key, value in metadata.items():
+    for key, value in dict(metadata).items():
         if potential_spend_tracking_metadata_substring in key:
             litellm_metadata[key] = value
     return litellm_metadata

tests/unit/litellm_core_utils/test_core_helpers.py

@@ -1,6 +1,8 @@
 """Tests for litellm_core_utils.core_helpers module."""
 
 import logging
+import threading
+from collections.abc import Iterator, Mapping
 from typing import Final
 
 import httpx
@@ -13,6 +15,7 @@ from litellm.litellm_core_utils.core_helpers import (
     budget_reservation_from_metadata,
     drop_params_env_flag,
     drop_params_flag,
+    get_litellm_metadata_from_kwargs,
     get_parent_otel_span_from_kwargs,
     get_or_create_metadata_bucket,
     get_provider_response_headers_from_hidden_params,
@@ -587,3 +590,55 @@ class TestProviderResponseHeadersInHiddenParams:
 
         assert get_provider_response_headers_from_hidden_params(sibling) is None
         assert "additional_headers" not in sibling._hidden_params
+
+
+def test_spend_metadata_merge_survives_the_logging_thread_inserting_into_the_request_metadata() -> None:
+    metadata: Final[dict[str, object]] = {}
+
+    def logging_thread_records_hidden_params() -> None:
+        metadata["hidden_params"] = {"response_cost": 0.1}
+
+    class KeyCheckedWhileTheLoggingThreadRuns(str):
+        def __contains__(self, key: str) -> bool:
+            writer: Final = threading.Thread(target=logging_thread_records_hidden_params)
+            writer.start()
+            writer.join()
+            return super().__contains__(key)
+
+    metadata[KeyCheckedWhileTheLoggingThreadRuns("user_api_key_alias")] = "alias"
+    metadata["user_api_key_hash"] = "hash"
+    metadata["requester_ip_address"] = "10.0.0.1"
+    litellm_metadata: Final[dict[str, object]] = {"user_api_key_alias": "stale", "model_group": "claude"}
+
+    merged: Final = get_litellm_metadata_from_kwargs(
+        {"litellm_params": {"metadata": metadata, "litellm_metadata": litellm_metadata}}
+    )
+
+    assert "hidden_params" in metadata
+    assert merged == {"user_api_key_alias": "alias", "user_api_key_hash": "hash", "model_group": "claude"}
+
+
+def test_spend_metadata_merge_reads_request_metadata_that_is_a_read_only_mapping() -> None:
+    class ReadOnlyMetadata(Mapping[str, object]):
+        def __init__(self, values: dict[str, object]) -> None:
+            self._values: Final = values
+
+        def __getitem__(self, key: str) -> object:
+            return self._values[key]
+
+        def __iter__(self) -> Iterator[str]:
+            return iter(self._values)
+
+        def __len__(self) -> int:
+            return len(self._values)
+
+    merged: Final = get_litellm_metadata_from_kwargs(
+        {
+            "litellm_params": {
+                "metadata": ReadOnlyMetadata({"user_api_key_hash": "hash", "requester_ip_address": "10.0.0.1"}),
+                "litellm_metadata": {"model_group": "claude"},
+            }
+        }
+    )
+
+    assert merged == {"user_api_key_hash": "hash", "model_group": "claude"}