fix(s3): replace colons in generated log filenames (#40452)

* fix(s3): replace colons in generated log filenames

Bedrock and Vertex AI batch file uploads use s3:// and gs:// URIs as
response ids. The shared filename sanitizer replaced slashes but kept
the scheme colon, producing log object keys that Hadoop-style consumers
reject as a relative path in an absolute URI.

Fixes #40234

* test(s3): drop docstrings flagged by review

---------

Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>

litellm/integrations/s3.py

@@ -277,7 +277,7 @@ def get_s3_object_key(
     start_time: datetime,
     s3_file_name: str,
 ) -> str:
-    sanitized_s3_file_name: Final = s3_file_name.replace("/", "_")
+    sanitized_s3_file_name: Final = s3_file_name.replace("/", "_").replace(":", "_")
     configured_prefix: Final = (s3_path.rstrip("/") + "/" if s3_path else "") + prefix
     date_segment: Final = start_time.strftime("%Y-%m-%d") + "/"
     # we need the s3 key to include the time, so we log cache hits too

tests/test_litellm/integrations/test_s3_v2.py

@@ -1273,9 +1273,7 @@ async def test_combined_prefix_reflects_in_s3_object_key():
     assert "myteam/apikey/" in key, f"Expected both prefixes in key: {key}"
 
 
-def test_s3_object_key_sanitizes_slashes_in_file_name():
-    """Response ids containing slashes (e.g. bedrock batch job ARNs) must not
-    create nested S3 folders; only path/prefix/date slashes are separators."""
+def test_s3_object_key_sanitizes_slashes_and_colons_in_file_name():
     from litellm.integrations.s3 import get_s3_object_key
 
     start_time = datetime(2026, 2, 11, 0, 35, 18, 391582)
@@ -1290,10 +1288,32 @@ def test_s3_object_key_sanitizes_slashes_in_file_name():
 
     assert key == (
         "LiteLLMAPPLogs/myteam/2026-02-11/"
-        "time-00-35-18-391582_arn:aws:bedrock:us-east-1:123456789012:model-invocation-job_gl18r6skk9yy.json"
+        "time-00-35-18-391582_arn_aws_bedrock_us-east-1_123456789012_model-invocation-job_gl18r6skk9yy.json"
     )
 
 
+@pytest.mark.parametrize(
+    "response_id",
+    [
+        "s3://example-batch-bucket/litellm-bedrock-files/input.jsonl",
+        "gs://example-batch-bucket/litellm-vertex-files/input.jsonl",
+    ],
+)
+def test_s3_object_key_has_no_colon_for_cloud_uri_file_ids(response_id: str):
+    from litellm.integrations.s3 import get_s3_object_key
+
+    key = get_s3_object_key(
+        s3_path="",
+        prefix="",
+        start_time=datetime(2026, 9, 7, 4, 51, 6, 685889),
+        s3_file_name=f"time-04-51-06-685889_{response_id}",
+    )
+
+    filename = key.rsplit("/", 1)[-1]
+    assert ":" not in filename
+    assert filename.endswith("_input.jsonl.json")
+
+
 def test_create_s3_batch_logging_element_flat_key_for_arn_response_id():
     """End-to-end through the s3_v2 element builder: an ARN response id must
     yield a flat file directly under the date segment."""