Python: Fix ai connector integration tests (#13311)

### Motivation and Context

<!-- Thank you for your contribution to the semantic-kernel repo!
Please help reviewers and future users, providing the following
information:
  1. Why is this change required?
  2. What problem does it solve?
  3. What scenario does it contribute to?
  4. If it fixes an open issue, please link to the issue here.
-->
Vertex AI and Bedrock integration tests have been failing.


### Description

<!-- Describe your changes, the overall approach, the underlying design.
These notes will help understanding how your code works. Thanks! -->
1. This PR fixes the vertex ai integration tests by upgrading the Google
AI SDK version.
2. This PR fixes the Bedrock integration tests by upgrading the models.


### Contribution Checklist

<!-- Before submitting this PR, please make sure: -->

- [x] The code builds clean without any errors or warnings
- [x] The PR follows the [SK Contribution
Guidelines](https://github.com/microsoft/semantic-kernel/blob/main/CONTRIBUTING.md)
and the [pre-submission formatting
script](https://github.com/microsoft/semantic-kernel/blob/main/CONTRIBUTING.md#development-scripts)
raises no violations
- [x] All unit tests pass, and I have added new tests where possible
- [x] I didn't break anyone 😄

---------

Co-authored-by: Copilot <[email protected]>
Co-authored-by: moonbox3 <[email protected]>

Author: 3 people

python/pyproject.toml

@@ -92,7 +92,7 @@ faiss = [
     "faiss-cpu>=1.10.0"
 ]
 google = [
-    "google-cloud-aiplatform == 1.97.0",
+    "google-cloud-aiplatform ~= 1.114.0",
     "google-generativeai ~= 0.8"
 ]
 hugging_face = [
@@ -154,7 +154,7 @@ usearch = [
     "pyarrow >= 12.0,< 22.0"
 ]
 weaviate = [
-    "weaviate-client>=4.10,<5.0,!=4.16.7",
+    "weaviate-client>=4.17.0,<5.0",
 ]
 
 [tool.uv]

python/semantic_kernel/connectors/ai/google/vertex_ai/services/utils.py

@@ -4,9 +4,8 @@
 import logging
 from typing import TYPE_CHECKING
 
-from google.cloud.aiplatform_v1beta1.types.content import Blob, Candidate, Part
-from google.cloud.aiplatform_v1beta1.types.tool import FunctionCall, FunctionResponse
-from vertexai.generative_models import FunctionDeclaration, Tool, ToolConfig
+from google.cloud.aiplatform_v1beta1.types.content import Candidate
+from vertexai.generative_models import FunctionDeclaration, Part, Tool, ToolConfig
 
 from semantic_kernel.connectors.ai.function_choice_behavior import FunctionChoiceType
 from semantic_kernel.connectors.ai.google.shared_utils import (
@@ -63,7 +62,7 @@ def format_user_message(message: ChatMessageContent) -> list[Part]:
     parts: list[Part] = []
     for item in message.items:
         if isinstance(item, TextContent):
-            parts.append(Part(text=message.content))
+            parts.append(Part.from_text(message.content))
         elif isinstance(item, ImageContent):
             parts.append(_create_image_part(item))
         else:
@@ -88,16 +87,15 @@ def format_assistant_message(message: ChatMessageContent) -> list[Part]:
     for item in message.items:
         if isinstance(item, TextContent):
             if item.text:
-                parts.append(Part(text=item.text))
+                parts.append(Part.from_text(item.text))
         elif isinstance(item, FunctionCallContent):
             parts.append(
-                Part(
-                    function_call=FunctionCall(
-                        name=item.name,
-                        # Convert the arguments to a dictionary if it is a string
-                        args=json.loads(item.arguments) if isinstance(item.arguments, str) else item.arguments,
-                    )
-                )
+                Part.from_dict({
+                    "function_call": {
+                        "name": item.name,
+                        "args": json.loads(item.arguments) if isinstance(item.arguments, str) else item.arguments,
+                    }
+                })
             )
         elif isinstance(item, ImageContent):
             parts.append(_create_image_part(item))
@@ -124,14 +122,11 @@ def format_tool_message(message: ChatMessageContent) -> list[Part]:
         if isinstance(item, FunctionResultContent):
             gemini_function_name = item.custom_fully_qualified_name(GEMINI_FUNCTION_NAME_SEPARATOR)
             parts.append(
-                Part(
-                    function_response=FunctionResponse(
-                        name=gemini_function_name,
-                        response={
-                            "name": gemini_function_name,
-                            "content": str(item.result),
-                        },
-                    )
+                Part.from_function_response(
+                    gemini_function_name,
+                    {
+                        "content": str(item.result),
+                    },
                 )
             )
 
@@ -177,7 +172,7 @@ def update_settings_from_function_choice_configuration(
 
 def _create_image_part(image_content: ImageContent) -> Part:
     if image_content.data_uri:
-        return Part(inline_data=Blob(mime_type=image_content.mime_type, data=image_content.data))
+        return Part.from_data(image_content.data, image_content.mime_type)  # type: ignore[arg-type]
 
     # The Google AI API doesn't support images from arbitrary URIs:
     # https://github.com/google-gemini/generative-ai-python/issues/357

python/semantic_kernel/connectors/ai/google/vertex_ai/services/vertex_ai_chat_completion.py

@@ -10,9 +10,8 @@
     from typing_extensions import override  # pragma: no cover
 
 import vertexai
-from google.cloud.aiplatform_v1beta1.types.content import Content
 from pydantic import ValidationError
-from vertexai.generative_models import Candidate, GenerationResponse, GenerativeModel
+from vertexai.generative_models import Candidate, Content, GenerationResponse, GenerativeModel
 
 from semantic_kernel.connectors.ai.chat_completion_client_base import ChatCompletionClientBase
 from semantic_kernel.connectors.ai.completion_usage import CompletionUsage
@@ -125,6 +124,7 @@ async def _inner_get_chat_message_contents(
         assert isinstance(settings, VertexAIChatPromptExecutionSettings)  # nosec
 
         vertexai.init(project=self.service_settings.project_id, location=self.service_settings.region)
+        assert self.service_settings.gemini_model_id is not None  # nosec
         model = GenerativeModel(
             self.service_settings.gemini_model_id,
             system_instruction=filter_system_message(chat_history),
@@ -154,6 +154,7 @@ async def _inner_get_streaming_chat_message_contents(
         assert isinstance(settings, VertexAIChatPromptExecutionSettings)  # nosec
 
         vertexai.init(project=self.service_settings.project_id, location=self.service_settings.region)
+        assert self.service_settings.gemini_model_id is not None  # nosec
         model = GenerativeModel(
             self.service_settings.gemini_model_id,
             system_instruction=filter_system_message(chat_history),

python/semantic_kernel/connectors/ai/google/vertex_ai/services/vertex_ai_text_completion.py

@@ -96,6 +96,7 @@ async def _inner_get_text_contents(
         assert isinstance(settings, VertexAITextPromptExecutionSettings)  # nosec
 
         vertexai.init(project=self.service_settings.project_id, location=self.service_settings.region)
+        assert self.service_settings.gemini_model_id is not None  # nosec
         model = GenerativeModel(self.service_settings.gemini_model_id)
 
         response: GenerationResponse = await model.generate_content_async(
@@ -117,6 +118,7 @@ async def _inner_get_streaming_text_contents(
         assert isinstance(settings, VertexAITextPromptExecutionSettings)  # nosec
 
         vertexai.init(project=self.service_settings.project_id, location=self.service_settings.region)
+        assert self.service_settings.gemini_model_id is not None  # nosec
         model = GenerativeModel(self.service_settings.gemini_model_id)
 
         response: AsyncIterable[GenerationResponse] = await model.generate_content_async(

python/semantic_kernel/connectors/ai/google/vertex_ai/services/vertex_ai_text_embedding.py

@@ -93,9 +93,10 @@ async def generate_raw_embeddings(
         assert isinstance(settings, VertexAIEmbeddingPromptExecutionSettings)  # nosec
 
         vertexai.init(project=self.service_settings.project_id, location=self.service_settings.region)
+        assert self.service_settings.embedding_model_id is not None  # nosec
         model = TextEmbeddingModel.from_pretrained(self.service_settings.embedding_model_id)
         response: list[TextEmbedding] = await model.get_embeddings_async(
-            texts,
+            texts,  # type: ignore[arg-type]
             **settings.prepare_settings_dict(),
         )
 

python/tests/integration/completions/chat_completion_test_base.py

@@ -155,8 +155,8 @@ def services(self) -> dict[str, tuple[ServiceType | None, type[PromptExecutionSe
                 OnnxGenAIChatCompletion(template=ONNXTemplate.PHI3V) if onnx_setup else None,
                 OnnxGenAIPromptExecutionSettings,
             ),
-            "bedrock_amazon_titan": (
-                self._try_create_bedrock_chat_completion_client("amazon.titan-text-premier-v1:0"),
+            "bedrock_amazon_nova": (
+                self._try_create_bedrock_chat_completion_client("amazon.nova-lite-v1:0"),
                 BedrockChatPromptExecutionSettings,
             ),
             "bedrock_ai21labs": (

python/tests/integration/completions/test_chat_completions.py

@@ -182,14 +182,14 @@ class Reasoning(KernelBaseModel):
         # endregion
         # region Bedrock
         pytest.param(
-            "bedrock_amazon_titan",
+            "bedrock_amazon_nova",
             {},
             [
                 ChatMessageContent(role=AuthorRole.USER, items=[TextContent(text="Hello")]),
                 ChatMessageContent(role=AuthorRole.USER, items=[TextContent(text="How are you today?")]),
             ],
             {},
-            id="bedrock_amazon_titan_text_input",
+            id="bedrock_amazon_nova_text_input",
         ),
         pytest.param(
             "bedrock_ai21labs",

python/tests/integration/completions/test_text_completion.py

@@ -258,7 +258,7 @@ def services(self) -> dict[str, tuple[ServiceType | None, type[PromptExecutionSe
             # Amazon Bedrock supports models from multiple providers but requests to and responses from the models are
             # inconsistent. So we need to test each model separately.
             "bedrock_amazon_titan": (
-                self._try_create_bedrock_text_completion_client("amazon.titan-text-premier-v1:0"),
+                self._try_create_bedrock_text_completion_client("amazon.titan-text-express-v1"),
                 BedrockTextPromptExecutionSettings,
             ),
             "bedrock_anthropic_claude": (

python/tests/unit/connectors/ai/google/vertex_ai/services/test_vertex_ai_chat_completion.py

@@ -4,8 +4,7 @@
 from unittest.mock import AsyncMock, patch
 
 import pytest
-from google.cloud.aiplatform_v1beta1.types.content import Content
-from vertexai.generative_models import GenerativeModel
+from vertexai.generative_models import Content, GenerativeModel
 
 from semantic_kernel.connectors.ai.function_choice_behavior import FunctionChoiceBehavior
 from semantic_kernel.connectors.ai.google.vertex_ai.services.vertex_ai_chat_completion import VertexAIChatCompletion
@@ -95,12 +94,23 @@ async def test_vertex_ai_chat_completion(
         chat_history, settings
     )
 
-    mock_vertex_ai_model_generate_content_async.assert_called_once_with(
-        contents=vertex_ai_chat_completion._prepare_chat_history_for_request(chat_history),
-        generation_config=settings.prepare_settings_dict(),
-        tools=None,
-        tool_config=None,
-    )
+    # Verify the call was made once
+    mock_vertex_ai_model_generate_content_async.assert_called_once()
+
+    # Get the actual call arguments
+    call_args = mock_vertex_ai_model_generate_content_async.call_args
+
+    # Verify the contents
+    contents = call_args.kwargs["contents"]
+    assert len(contents) == 1
+    assert contents[0].role == "user"
+    assert len(contents[0].parts) == 1
+    assert contents[0].parts[0].text == "test_prompt"
+
+    # Verify other arguments
+    assert call_args.kwargs["generation_config"] == settings.prepare_settings_dict()
+    assert call_args.kwargs["tools"] is None
+    assert call_args.kwargs["tool_config"] is None
     assert len(responses) == 1
     assert responses[0].role == "assistant"
     assert responses[0].content == mock_vertex_ai_chat_completion_response.candidates[0].content.parts[0].text
@@ -188,12 +198,23 @@ async def test_vertex_ai_chat_completion_with_function_choice_behavior_no_tool_c
         kernel=kernel,
     )
 
-    mock_vertex_ai_model_generate_content_async.assert_awaited_once_with(
-        contents=vertex_ai_chat_completion._prepare_chat_history_for_request(chat_history),
-        generation_config=settings.prepare_settings_dict(),
-        tools=None,
-        tool_config=None,
-    )
+    # Verify the call was made once
+    mock_vertex_ai_model_generate_content_async.assert_awaited_once()
+
+    # Get the actual call arguments
+    call_args = mock_vertex_ai_model_generate_content_async.await_args
+
+    # Verify the contents
+    contents = call_args.kwargs["contents"]
+    assert len(contents) == 1
+    assert contents[0].role == "user"
+    assert len(contents[0].parts) == 1
+    assert contents[0].parts[0].text == "test_prompt"
+
+    # Verify other arguments
+    assert call_args.kwargs["generation_config"] == settings.prepare_settings_dict()
+    assert call_args.kwargs["tools"] is None
+    assert call_args.kwargs["tool_config"] is None
     assert len(responses) == 1
     assert responses[0].role == "assistant"
     assert responses[0].content == mock_vertex_ai_chat_completion_response.candidates[0].content.parts[0].text
@@ -225,13 +246,24 @@ async def test_vertex_ai_streaming_chat_completion(
         assert "usage" in messages[0].metadata
         assert "prompt_feedback" in messages[0].metadata
 
-    mock_vertex_ai_model_generate_content_async.assert_called_once_with(
-        contents=vertex_ai_chat_completion._prepare_chat_history_for_request(chat_history),
-        generation_config=settings.prepare_settings_dict(),
-        tools=None,
-        tool_config=None,
-        stream=True,
-    )
+    # Verify the call was made once
+    mock_vertex_ai_model_generate_content_async.assert_called_once()
+
+    # Get the actual call arguments
+    call_args = mock_vertex_ai_model_generate_content_async.call_args
+
+    # Verify the contents
+    contents = call_args.kwargs["contents"]
+    assert len(contents) == 1
+    assert contents[0].role == "user"
+    assert len(contents[0].parts) == 1
+    assert contents[0].parts[0].text == "test_prompt"
+
+    # Verify other arguments
+    assert call_args.kwargs["generation_config"] == settings.prepare_settings_dict()
+    assert call_args.kwargs["tools"] is None
+    assert call_args.kwargs["tool_config"] is None
+    assert call_args.kwargs["stream"] is True
 
 
 async def test_vertex_ai_streaming_chat_completion_with_function_choice_behavior_fail_verification(
@@ -328,13 +360,24 @@ async def test_vertex_ai_streaming_chat_completion_with_function_choice_behavior
         assert len(messages) == 1
         assert messages[0].role == "assistant"
 
-    mock_vertex_ai_model_generate_content_async.assert_awaited_once_with(
-        contents=vertex_ai_chat_completion._prepare_chat_history_for_request(chat_history),
-        generation_config=settings.prepare_settings_dict(),
-        tools=None,
-        tool_config=None,
-        stream=True,
-    )
+    # Verify the call was made once
+    mock_vertex_ai_model_generate_content_async.assert_awaited_once()
+
+    # Get the actual call arguments
+    call_args = mock_vertex_ai_model_generate_content_async.await_args
+
+    # Verify the contents
+    contents = call_args.kwargs["contents"]
+    assert len(contents) == 1
+    assert contents[0].role == "user"
+    assert len(contents[0].parts) == 1
+    assert contents[0].parts[0].text == "test_prompt"
+
+    # Verify other arguments
+    assert call_args.kwargs["generation_config"] == settings.prepare_settings_dict()
+    assert call_args.kwargs["tools"] is None
+    assert call_args.kwargs["tool_config"] is None
+    assert call_args.kwargs["stream"] is True
 
 
 # endregion streaming chat completion

python/tests/unit/connectors/ai/google/vertex_ai/services/test_vertex_ai_utils.py

@@ -1,7 +1,8 @@
 # Copyright (c) Microsoft. All rights reserved.
 
 import pytest
-from google.cloud.aiplatform_v1beta1.types.content import Candidate, Part
+from google.cloud.aiplatform_v1beta1.types.content import Candidate
+from vertexai.generative_models import Part
 
 from semantic_kernel.connectors.ai.google.vertex_ai.services.utils import (
     finish_reason_from_vertex_ai_to_semantic_kernel,