fix(prompt_management): don't route requests without prompt_id to prompt managers that can't run them

UI-injected empty vector_store_ids/tags/guardrails on a DB model tripped the
dynamic-param check, and the prompt-management fallback then handed the request
to the first registered prompt manager (e.g. a saved dotprompt), whose sync
path raised "prompt_id is required" as a 500 on every /chat/completions call.

Empty dynamic params no longer count as a trigger, the fallback skips managers
whose should_run_prompt_management declines a None prompt_id, and the sync base
path returns the request unchanged for a None prompt_id like the async path.

litellm/integrations/prompt_management_base.py

@@ -165,7 +165,7 @@ class PromptManagementBase(ABC):
         ignore_prompt_manager_optional_params: bool | None = False,
     ) -> tuple[str, list[AllMessageValues], dict]:
         if prompt_id is None:
-            raise ValueError("prompt_id is required for Prompt Management Base class")
+            return model, messages, non_default_params
         if not self.should_run_prompt_management(
             prompt_id=prompt_id,
             prompt_spec=prompt_spec,

litellm/litellm_core_utils/litellm_logging.py

@@ -832,8 +832,8 @@ class Logging(LiteLLMLoggingBaseClass):
 
         eg. AnthropicCacheControlHook and BedrockKnowledgeBaseHook both don't require a `prompt_id` to be passed in, they are triggered by dynamic params
         """
-        for param in non_default_params:
-            if param in DynamicPromptManagementParamLiteral.list_all_params():
+        for param in DynamicPromptManagementParamLiteral.list_all_params():
+            if non_default_params.get(param):
                 return True
 
         #############################################################################
@@ -966,6 +966,23 @@ class Logging(LiteLLMLoggingBaseClass):
 
         return None
 
+    @staticmethod
+    def _prompt_manager_runs_without_prompt_id(
+        logger: CustomLogger,
+        prompt_spec: PromptSpec | None,
+        dynamic_callback_params: StandardCallbackDynamicParams | None,
+    ) -> bool:
+        if not isinstance(logger, CustomPromptManagement):
+            return False
+        try:
+            return logger.should_run_prompt_management(
+                prompt_id=None,
+                prompt_spec=prompt_spec,
+                dynamic_callback_params=dynamic_callback_params or StandardCallbackDynamicParams(),
+            )
+        except Exception:
+            return False
+
     def get_custom_logger_for_prompt_management(
         self,
         model: str,
@@ -1016,8 +1033,13 @@ class Logging(LiteLLMLoggingBaseClass):
             callback_type=CustomPromptManagement
         )
 
-        if prompt_management_loggers:
-            logger: Final = prompt_management_loggers[0]
+        for logger in prompt_management_loggers:
+            if prompt_id is None and not self._prompt_manager_runs_without_prompt_id(
+                logger=logger,
+                prompt_spec=prompt_spec,
+                dynamic_callback_params=dynamic_callback_params,
+            ):
+                continue
             self.model_call_details["prompt_integration"] = logger.__class__.__name__
             return logger
 

tests/test_litellm/litellm_core_utils/test_litellm_logging.py

@@ -5105,3 +5105,81 @@ def test_set_cost_breakdown_stores_vertex_location():
         cost_for_built_in_tools_cost_usd_dollar=0.0,
     )
     assert no_location.cost_breakdown.get("vertex_location") is None
+
+
+def test_prompt_hooks_skip_prompt_managers_when_no_prompt_id(logging_obj, tmp_path, monkeypatch):
+    """
+    Regression for UI-injected `vector_store_ids: []` and always-on non-empty `vector_store_ids`
+    with a registered prompt manager (e.g. dotprompt): requests without a prompt_id 500'd with
+    "prompt_id is required for Prompt Management Base class" instead of completing normally.
+    """
+    from litellm.integrations.dotprompt.dotprompt_manager import DotpromptManager
+    from litellm.integrations.vector_store_integrations.vector_store_pre_call_hook import (
+        VectorStorePreCallHook,
+    )
+    from litellm.types.vector_stores import LiteLLM_ManagedVectorStore
+    from litellm.vector_stores.vector_store_registry import VectorStoreRegistry
+
+    (tmp_path / "stem.prompt").write_text("---\nmodel: gemini-2.5-flash\n---\nyou are a stem tutor\n")
+    dotprompt_manager = DotpromptManager(prompt_directory=str(tmp_path))
+    litellm.logging_callback_manager.add_litellm_callback(dotprompt_manager)
+    monkeypatch.setattr(
+        litellm,
+        "vector_store_registry",
+        VectorStoreRegistry(
+            vector_stores=[LiteLLM_ManagedVectorStore(vector_store_id="vs_123", custom_llm_provider="openai")]
+        ),
+    )
+
+    messages = [{"role": "user", "content": "hi"}]
+    try:
+        assert not logging_obj.should_run_prompt_management_hooks(
+            prompt_id=None, non_default_params={"vector_store_ids": []}
+        )
+
+        assert logging_obj.get_chat_completion_prompt(
+            model="gemini-2.5-flash",
+            messages=messages,
+            non_default_params={"vector_store_ids": []},
+            prompt_variables=None,
+            prompt_id=None,
+        ) == ("gemini-2.5-flash", messages, {"vector_store_ids": []})
+
+        assert dotprompt_manager.get_chat_completion_prompt(
+            model="gemini-2.5-flash",
+            messages=messages,
+            non_default_params={},
+            prompt_id=None,
+            prompt_variables=None,
+            dynamic_callback_params={},
+        ) == ("gemini-2.5-flash", messages, {})
+
+        assert logging_obj.should_run_prompt_management_hooks(
+            prompt_id=None, non_default_params={"vector_store_ids": ["vs_123"]}
+        )
+        assert isinstance(
+            logging_obj.get_custom_logger_for_prompt_management(
+                model="gemini-2.5-flash",
+                non_default_params={"vector_store_ids": ["vs_123"]},
+                prompt_id=None,
+                dynamic_callback_params={},
+            ),
+            VectorStorePreCallHook,
+        )
+
+        assert isinstance(
+            logging_obj.get_custom_logger_for_prompt_management(
+                model="gemini-2.5-flash",
+                non_default_params={},
+                prompt_id="stem",
+                dynamic_callback_params={},
+            ),
+            DotpromptManager,
+        )
+    finally:
+        litellm.logging_callback_manager.remove_callback_from_list_by_object(litellm.callbacks, dotprompt_manager)
+        litellm.logging_callback_manager.remove_callback_from_list_by_object(
+            litellm._async_success_callback, dotprompt_manager
+        )
+        for hook in [cb for cb in litellm.callbacks if isinstance(cb, VectorStorePreCallHook)]:
+            litellm.logging_callback_manager.remove_callback_from_list_by_object(litellm.callbacks, hook)