fix: resolve AttributeError by supporting both Pydantic and Protobuf AgentCard serialization

PiperOrigin-RevId: 930002178

Author: vertex-sdk-bot

setup.py

@@ -280,6 +280,7 @@
         "mock",
         "pytest-xdist",
         "Pillow",
+        "a2a-sdk",
         "scikit-learn<1.6.0; python_version<='3.10'",
         "scikit-learn; python_version>'3.10'",
         # Lazy import requires > 2.12.0

tests/unit/vertex_a2a/test_agent_engines_a2a.py

@@ -0,0 +1,301 @@
+# Copyright 2026 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+import os
+import sys
+import tempfile
+from unittest import mock
+import pytest
+import cloudpickle
+import pydantic
+
+from google import auth
+from google.api_core import operation as ga_operation
+from google.auth import credentials as auth_credentials
+from google.cloud import storage
+from google.cloud import aiplatform
+from google.cloud.aiplatform import base
+
+from google.cloud.aiplatform_v1 import types
+from google.cloud.aiplatform_v1.services import reasoning_engine_service
+from vertexai import agent_engines
+from vertexai.agent_engines import _agent_engines
+from vertexai.agent_engines import _utils
+from google.protobuf import struct_pb2
+
+
+class CapitalizeEngine:
+    """A sample Agent Engine."""
+
+    def query(self, unused_arbitrary_string_name: str) -> str:
+        """Runs the engine."""
+        return unused_arbitrary_string_name.upper()
+
+
+class CapitalizeEngineWithCard(CapitalizeEngine):
+
+    def __init__(self, card):
+        self.agent_card = card
+
+    def __getstate__(self):
+        state = self.__dict__.copy()
+        if hasattr(self.agent_card, "DESCRIPTOR"):
+            state["agent_card"] = None
+        return state
+
+    def __setstate__(self, state):
+        self.__dict__.update(state)
+
+
+class DummyPydanticCard(pydantic.BaseModel):
+    name: str = "test_pydantic_card"
+
+
+def _create_empty_fake_package(package_name: str) -> str:
+    temp_dir = tempfile.mkdtemp()
+    package_dir = os.path.join(temp_dir, package_name)
+    os.makedirs(package_dir)
+    init_path = os.path.join(package_dir, "__init__.py")
+    open(init_path, "w").close()
+    return temp_dir
+
+
+_TEST_CREDENTIALS = mock.Mock(spec=auth_credentials.AnonymousCredentials())
+_TEST_STAGING_BUCKET = "gs://test-bucket"
+_TEST_LOCATION = "us-central1"
+_TEST_PROJECT = "test-project"
+_TEST_RESOURCE_ID = "1028944691210842416"
+_TEST_PARENT = f"projects/{_TEST_PROJECT}/locations/{_TEST_LOCATION}"
+_TEST_AGENT_ENGINE_RESOURCE_NAME = (
+    f"{_TEST_PARENT}/reasoningEngines/{_TEST_RESOURCE_ID}"
+)
+_TEST_AGENT_ENGINE_DISPLAY_NAME = "Agent Engine Display Name"
+_TEST_GCS_DIR_NAME = _agent_engines._DEFAULT_GCS_DIR_NAME
+_TEST_BLOB_FILENAME = _agent_engines._BLOB_FILENAME
+_TEST_REQUIREMENTS_FILE = _agent_engines._REQUIREMENTS_FILE
+_TEST_EXTRA_PACKAGES_FILE = _agent_engines._EXTRA_PACKAGES_FILE
+_TEST_STANDARD_API_MODE = _agent_engines._STANDARD_API_MODE
+_TEST_DEFAULT_METHOD_NAME = _agent_engines._DEFAULT_METHOD_NAME
+_TEST_MODE_KEY_IN_SCHEMA = _agent_engines._MODE_KEY_IN_SCHEMA
+
+_TEST_AGENT_ENGINE_EXTRA_PACKAGE = "fake.py"
+
+_TEST_AGENT_ENGINE_EXTRA_PACKAGE_PATH = _create_empty_fake_package(
+    _TEST_AGENT_ENGINE_EXTRA_PACKAGE
+)
+
+_TEST_AGENT_ENGINE_REQUIREMENTS = [
+    "google-cloud-aiplatform==1.29.0",
+    "langchain",
+]
+
+_TEST_AGENT_ENGINE_GCS_URI = "{}/{}/{}".format(
+    _TEST_STAGING_BUCKET,
+    _TEST_GCS_DIR_NAME,
+    _TEST_BLOB_FILENAME,
+)
+_TEST_AGENT_ENGINE_DEPENDENCY_FILES_GCS_URI = "{}/{}/{}".format(
+    _TEST_STAGING_BUCKET,
+    _TEST_GCS_DIR_NAME,
+    _TEST_EXTRA_PACKAGES_FILE,
+)
+_TEST_AGENT_ENGINE_REQUIREMENTS_GCS_URI = "{}/{}/{}".format(
+    _TEST_STAGING_BUCKET,
+    _TEST_GCS_DIR_NAME,
+    _TEST_REQUIREMENTS_FILE,
+)
+
+_TEST_AGENT_ENGINE_QUERY_SCHEMA = _utils.to_proto(
+    _utils.generate_schema(
+        CapitalizeEngine().query,
+        schema_name=_TEST_DEFAULT_METHOD_NAME,
+    )
+)
+_TEST_AGENT_ENGINE_QUERY_SCHEMA[_TEST_MODE_KEY_IN_SCHEMA] = _TEST_STANDARD_API_MODE
+
+_TEST_AGENT_ENGINE_PACKAGE_SPEC = types.ReasoningEngineSpec.PackageSpec(
+    python_version=f"{sys.version_info.major}.{sys.version_info.minor}",
+    pickle_object_gcs_uri=_TEST_AGENT_ENGINE_GCS_URI,
+    dependency_files_gcs_uri=_TEST_AGENT_ENGINE_DEPENDENCY_FILES_GCS_URI,
+    requirements_gcs_uri=_TEST_AGENT_ENGINE_REQUIREMENTS_GCS_URI,
+)
+
+_TEST_AGENT_ENGINE_OBJ = types.ReasoningEngine(
+    name=_TEST_AGENT_ENGINE_RESOURCE_NAME,
+    spec=types.ReasoningEngineSpec(
+        package_spec=_TEST_AGENT_ENGINE_PACKAGE_SPEC,
+        agent_framework=_agent_engines._DEFAULT_AGENT_FRAMEWORK,
+    ),
+)
+_TEST_AGENT_ENGINE_OBJ.spec.class_methods.append(_TEST_AGENT_ENGINE_QUERY_SCHEMA)
+
+
+@pytest.fixture(scope="module")
+def google_auth_mock():
+    with mock.patch.object(auth, "default") as google_auth_mock:
+        google_auth_mock.return_value = (
+            auth_credentials.AnonymousCredentials(),
+            _TEST_PROJECT,
+        )
+        yield google_auth_mock
+
+
+@pytest.fixture(scope="module")
+def cloud_storage_create_bucket_mock():
+    with mock.patch.object(storage, "Client") as cloud_storage_mock:
+        bucket_mock = mock.Mock(spec=storage.Bucket)
+        bucket_mock.blob.return_value.open.return_value = "blob_file"
+        bucket_mock.blob.return_value.upload_from_filename.return_value = None
+        bucket_mock.blob.return_value.upload_from_string.return_value = None
+
+        cloud_storage_mock.get_bucket = mock.Mock(
+            side_effect=ValueError("bucket not found")
+        )
+        cloud_storage_mock.bucket.return_value = bucket_mock
+        cloud_storage_mock.create_bucket.return_value = bucket_mock
+
+        yield cloud_storage_mock
+
+
+@pytest.fixture(scope="module")
+def cloudpickle_load_mock():
+    with mock.patch.object(cloudpickle, "load") as cloudpickle_load_mock:
+        yield cloudpickle_load_mock
+
+
+@pytest.fixture(scope="module")
+def create_agent_engine_mock():
+    with mock.patch.object(
+        reasoning_engine_service.ReasoningEngineServiceClient,
+        "create_reasoning_engine",
+    ) as create_agent_engine_mock:
+        create_agent_engine_lro_mock = mock.Mock(spec=ga_operation.Operation)
+        create_agent_engine_lro_mock.result.return_value = _TEST_AGENT_ENGINE_OBJ
+        create_agent_engine_mock.return_value = create_agent_engine_lro_mock
+        yield create_agent_engine_mock
+
+
+@pytest.fixture(scope="function")
+def get_gca_resource_mock():
+    with mock.patch.object(
+        base.VertexAiResourceNoun,
+        "_get_gca_resource",
+    ) as get_gca_resource_mock:
+        get_gca_resource_mock.return_value = _TEST_AGENT_ENGINE_OBJ
+        yield get_gca_resource_mock
+
+
+@pytest.mark.usefixtures("google_auth_mock")
+class TestAgentEngineA2A:
+    def setup_method(self):
+        aiplatform.init(
+            project=_TEST_PROJECT,
+            location=_TEST_LOCATION,
+            credentials=_TEST_CREDENTIALS,
+            staging_bucket=_TEST_STAGING_BUCKET,
+        )
+
+    def test_create_agent_engine_with_protobuf_agent_card(
+        self,
+        create_agent_engine_mock,
+        cloud_storage_create_bucket_mock,
+        cloudpickle_load_mock,
+        get_gca_resource_mock,
+    ):
+        from a2a.compat.v0_3 import a2a_v0_3_pb2 as a2a_pb2
+
+        card = a2a_pb2.AgentCard(name="test_agent_card")
+        agent = CapitalizeEngineWithCard(card)
+
+        agent_engines.create(
+            agent,
+            display_name=_TEST_AGENT_ENGINE_DISPLAY_NAME,
+            requirements=_TEST_AGENT_ENGINE_REQUIREMENTS,
+            extra_packages=[_TEST_AGENT_ENGINE_EXTRA_PACKAGE_PATH],
+        )
+
+        expected_reasoning_engine = types.ReasoningEngine(
+            display_name=_TEST_AGENT_ENGINE_DISPLAY_NAME,
+            spec=types.ReasoningEngineSpec(
+                package_spec=_TEST_AGENT_ENGINE_PACKAGE_SPEC,
+                agent_framework=_agent_engines._DEFAULT_AGENT_FRAMEWORK,
+            ),
+        )
+        from google.protobuf import json_format
+
+        expected_class_method = struct_pb2.Struct()
+        expected_class_method.CopyFrom(_TEST_AGENT_ENGINE_QUERY_SCHEMA)
+        expected_class_method["a2a_agent_card"] = json_format.MessageToJson(card)
+        expected_reasoning_engine.spec.class_methods.append(expected_class_method)
+
+        create_agent_engine_mock.assert_called_with(
+            parent=_TEST_PARENT,
+            reasoning_engine=expected_reasoning_engine,
+        )
+
+    def test_create_agent_engine_with_pydantic_agent_card(
+        self,
+        create_agent_engine_mock,
+        cloud_storage_create_bucket_mock,
+        cloudpickle_load_mock,
+        get_gca_resource_mock,
+    ):
+        card = DummyPydanticCard()
+        agent = CapitalizeEngineWithCard(card)
+
+        agent_engines.create(
+            agent,
+            display_name=_TEST_AGENT_ENGINE_DISPLAY_NAME,
+            requirements=_TEST_AGENT_ENGINE_REQUIREMENTS,
+            extra_packages=[_TEST_AGENT_ENGINE_EXTRA_PACKAGE_PATH],
+        )
+
+        expected_reasoning_engine = types.ReasoningEngine(
+            display_name=_TEST_AGENT_ENGINE_DISPLAY_NAME,
+            spec=types.ReasoningEngineSpec(
+                package_spec=_TEST_AGENT_ENGINE_PACKAGE_SPEC,
+                agent_framework=_agent_engines._DEFAULT_AGENT_FRAMEWORK,
+            ),
+        )
+
+        expected_class_method = struct_pb2.Struct()
+        expected_class_method.CopyFrom(_TEST_AGENT_ENGINE_QUERY_SCHEMA)
+        expected_class_method["a2a_agent_card"] = card.model_dump_json()
+        expected_reasoning_engine.spec.class_methods.append(expected_class_method)
+
+        create_agent_engine_mock.assert_called_with(
+            parent=_TEST_PARENT,
+            reasoning_engine=expected_reasoning_engine,
+        )
+
+    def test_create_agent_engine_with_invalid_agent_card(
+        self,
+        create_agent_engine_mock,
+        cloud_storage_create_bucket_mock,
+        cloudpickle_load_mock,
+        get_gca_resource_mock,
+    ):
+        agent = CapitalizeEngineWithCard(card="invalid_card_type_string")
+
+        with pytest.raises(
+            TypeError,
+            match="Unsupported AgentCard type",
+        ):
+            agent_engines.create(
+                agent,
+                display_name=_TEST_AGENT_ENGINE_DISPLAY_NAME,
+                requirements=_TEST_AGENT_ENGINE_REQUIREMENTS,
+                extra_packages=[_TEST_AGENT_ENGINE_EXTRA_PACKAGE_PATH],
+            )

vertexai/_genai/_agent_engines_utils.py

@@ -652,10 +652,9 @@ def _generate_class_methods_spec_or_raise(
 
             class_method = _to_proto(schema_dict)
             class_method[_MODE_KEY_IN_SCHEMA] = mode
-            if hasattr(agent, "agent_card"):
-                class_method[_A2A_AGENT_CARD] = json_format.MessageToJson(
-                    getattr(agent, "agent_card")
-                )
+            card = getattr(agent, "agent_card", None)
+            if card is not None:
+                class_method[_A2A_AGENT_CARD] = _serialize_agent_card_to_json(card)
             class_methods_spec.append(class_method)
 
     return class_methods_spec
@@ -2148,3 +2147,59 @@ def _add_telemetry_enablement_env(
         return env_vars
 
     return env_vars | env_to_add
+
+
+def _serialize_agent_card_to_dict(card: Any) -> Optional[Dict[str, Any]]:
+    """Validates and serializes an AgentCard to a dictionary representation.
+
+    Args:
+        card: The AgentCard instance (Pydantic model or Protobuf Message).
+
+    Returns:
+        The serialized card as a dictionary.
+
+    Raises:
+        TypeError: If the card type is not supported.
+    """
+    if card is None:
+        return None
+
+    if hasattr(card, "model_dump"):
+        return typing.cast(dict[str, Any], card.model_dump(exclude_none=True))
+    elif hasattr(card, "DESCRIPTOR"):
+        from google.protobuf import json_format
+
+        return typing.cast(dict[str, Any], json_format.MessageToDict(card))
+    else:
+        raise TypeError(
+            f"Unsupported AgentCard type: {type(card)}. "
+            "Only Pydantic models and Protobuf Messages are supported."
+        )
+
+
+def _serialize_agent_card_to_json(card: Any) -> Optional[str]:
+    """Validates and serializes an AgentCard to a JSON string representation.
+
+    Args:
+        card: The AgentCard instance (Pydantic model or Protobuf Message).
+
+    Returns:
+        The serialized card as a JSON string.
+
+    Raises:
+        TypeError: If the card type is not supported.
+    """
+    if card is None:
+        return None
+
+    if hasattr(card, "model_dump_json"):
+        return typing.cast(str, card.model_dump_json())
+    elif hasattr(card, "DESCRIPTOR"):
+        from google.protobuf import json_format
+
+        return typing.cast(str, json_format.MessageToJson(card))
+    else:
+        raise TypeError(
+            f"Unsupported AgentCard type: {type(card)}. "
+            "Only Pydantic models and Protobuf Messages are supported."
+        )