Merge 8bd4d59ffc into 587f227b9d

litellm/litellm_core_utils/litellm_logging.py

@@ -1042,7 +1042,7 @@ class Logging(LiteLLMLoggingBaseClass):
         )
 
         for logger in prompt_management_loggers:
-            if prompt_id is None and not self._prompt_manager_runs_without_prompt_id(
+            if not prompt_id and not self._prompt_manager_runs_without_prompt_id(
                 logger=logger,
                 prompt_spec=prompt_spec,
                 dynamic_callback_params=dynamic_callback_params,

tests/test_litellm/litellm_core_utils/test_litellm_logging.py

@@ -5973,3 +5973,34 @@ def test_failure_handler_helper_fn_builds_payload_once_per_exception():
     other_exc = _raise_and_catch(_ClientError(status_code=429, message="rate limited"))
     obj._failure_handler_helper_fn(exception=other_exc, traceback_exception="")
     assert obj.model_call_details["standard_logging_object"] is not first_payload
+
+
+def test_prompt_hooks_skip_prompt_managers_when_prompt_id_is_empty(logging_obj, tmp_path, monkeypatch):
+    """
+    `should_run_prompt_management_hooks` gates on `if prompt_id`, so an empty string reaches
+    dispatch as a call that names no prompt. Dispatch gated on `prompt_id is None`, so it
+    handed the call to the dotprompt manager anyway and the cache_control hook never ran.
+    """
+    from litellm.integrations.dotprompt.dotprompt_manager import DotpromptManager
+
+    (tmp_path / "stem.prompt").write_text("---\nmodel: claude-opus-4-5\n---\nyou are a stem tutor\n")
+    dotprompt_manager = DotpromptManager(prompt_directory=str(tmp_path))
+    monkeypatch.setattr(litellm, "callbacks", [dotprompt_manager])
+
+    cache_control_params = {"cache_control_injection_points": [{"role": "system", "location": "message"}]}
+    messages = [
+        {"role": "system", "content": "you are a stem tutor"},
+        {"role": "user", "content": "hi"},
+    ]
+
+    assert logging_obj.should_run_prompt_management_hooks(prompt_id="", non_default_params=cache_control_params)
+
+    _, compiled_messages, _ = logging_obj.get_chat_completion_prompt(
+        model="claude-opus-4-5",
+        messages=messages,
+        non_default_params=cache_control_params,
+        prompt_variables=None,
+        prompt_id="",
+    )
+
+    assert compiled_messages[0]["cache_control"] == {"type": "ephemeral"}