fix(bedrock/converse): strip tool_choice from inference_params when no tools present

When a client (e.g. n8n) sends tool_choice unconditionally but
the message has no tools, Bedrock throws ValidationException because
tool_choice is passed through to inferenceConfig.

Add an else branch that pops tool_choice from inference_params
when bedrock_tools is empty, consistent with how the if-branch
already pops it when tools ARE present.

litellm/llms/bedrock/chat/converse_transformation.py

@@ -1519,6 +1519,8 @@ class AmazonConverseConfig(BaseConfig):
             )
             if tool_choice_values is not None:
                 bedrock_tool_config["toolChoice"] = tool_choice_values
+        else:
+            inference_params.pop("tool_choice", None)
 
         data: CommonRequestObject = {
             "additionalModelRequestFields": additional_request_params,

tests/test_litellm/llms/bedrock/chat/test_converse_transformation.py

@@ -4578,3 +4578,42 @@ def test_transform_response_does_not_leak_body_on_parse_failure():
     msg = str(exc_info.value)
     assert "secret content" not in msg
     assert "Error converting to valid response block" in msg
+
+
+def test_tool_choice_stripped_when_no_tools():
+    """
+    When bedrock_tools is empty but inference_params contains tool_choice,
+    it must be popped before _transform_inference_params to avoid
+    ValidationException from Bedrock.
+    """
+    from unittest.mock import MagicMock, patch
+
+    from litellm.llms.bedrock.chat.converse_transformation import (
+        AmazonConverseConfig,
+    )
+
+    config = AmazonConverseConfig()
+    inference_params = {
+        "temperature": 0.7,
+        "tool_choice": {"auto": {}},
+    }
+
+    with patch.object(
+        config,
+        "_transform_request",
+        wraps=config._transform_request,
+    ):
+        messages = [{"role": "user", "content": [{"text": "Hello"}]}]
+        optional_params = dict(inference_params)
+
+        result = config.transform_request(
+            model="anthropic.claude-3-sonnet-20240229-v1:0",
+            messages=[{"role": "user", "content": "Hello"}],
+            optional_params=optional_params,
+            litellm_params={"api_base": None},
+            headers={},
+        )
+
+        inference_config = result.get("inferenceConfig", {})
+        assert "tool_choice" not in inference_config
+        assert result.get("toolConfig") is None