chore: import upstream snapshot with attribution

python/tests/unit/connectors/ai/onnx/conftest.py

@@ -0,0 +1,32 @@
+# Copyright (c) Microsoft. All rights reserved.
+from pytest import fixture
+
+
+@fixture()
+def onnx_unit_test_env(monkeypatch, exclude_list, override_env_param_dict):
+    """Fixture to set environment variables for OnnxGenAISettings."""
+    if exclude_list is None:
+        exclude_list = []
+
+    if override_env_param_dict is None:
+        override_env_param_dict = {}
+
+    env_vars = {
+        "ONNX_GEN_AI_CHAT_MODEL_FOLDER": "test",
+        "ONNX_GEN_AI_TEXT_MODEL_FOLDER": "test",
+    }
+
+    env_vars.update(override_env_param_dict)
+
+    for key, value in env_vars.items():
+        if key not in exclude_list:
+            monkeypatch.setenv(key, value)
+        else:
+            monkeypatch.delenv(key, raising=False)
+
+    return env_vars
+
+
+gen_ai_config = {"model": {"test": "test"}}
+
+gen_ai_config_vision = {"model": {"vision": "test"}}

python/tests/unit/connectors/ai/onnx/services/test_onnx_chat_completion.py

@@ -0,0 +1,187 @@
+# Copyright (c) Microsoft. All rights reserved.
+import json
+import os
+from unittest.mock import MagicMock, mock_open, patch
+
+import pytest
+
+from semantic_kernel.connectors.ai.onnx import OnnxGenAIChatCompletion, OnnxGenAIPromptExecutionSettings, ONNXTemplate
+from semantic_kernel.contents import AuthorRole, ChatHistory, ChatMessageContent, ImageContent
+from semantic_kernel.exceptions import ServiceInitializationError, ServiceInvalidExecutionSettingsError
+from semantic_kernel.kernel import Kernel
+from tests.unit.connectors.ai.onnx.conftest import gen_ai_config, gen_ai_config_vision
+
+
+@patch("builtins.open", new_callable=mock_open, read_data=json.dumps(gen_ai_config))
+@patch("onnxruntime_genai.Model")
+@patch("onnxruntime_genai.Tokenizer")
+def test_onnx_chat_completion_with_valid_env_variable(gen_ai_config, model, tokenizer, onnx_unit_test_env):
+    service = OnnxGenAIChatCompletion(template=ONNXTemplate.PHI3, env_file_path="test.env")
+    assert not service.enable_multi_modality
+
+
+@patch("builtins.open", new_callable=mock_open, read_data=json.dumps(gen_ai_config_vision))
+@patch("onnxruntime_genai.Model")
+@patch("onnxruntime_genai.Tokenizer")
+def test_onnx_chat_completion_with_vision_valid_env_variable(
+    gen_ai_vision_config, model, tokenizer, onnx_unit_test_env
+):
+    service = OnnxGenAIChatCompletion(template=ONNXTemplate.PHI3, env_file_path="test.env")
+    assert service.enable_multi_modality
+
+
+@patch("builtins.open", new_callable=mock_open, read_data=json.dumps(gen_ai_config))
+@patch("onnxruntime_genai.Model")
+@patch("onnxruntime_genai.Tokenizer")
+def test_onnx_chat_completion_with_valid_parameter(gen_ai_config, model, tokenizer):
+    assert OnnxGenAIChatCompletion(ai_model_path="/valid_path", template=ONNXTemplate.PHI3)
+
+
+@patch("builtins.open", new_callable=mock_open, read_data=json.dumps(gen_ai_config))
+@patch("onnxruntime_genai.Model")
+@patch("onnxruntime_genai.Tokenizer")
+def test_onnx_chat_completion_with_str_template(gen_ai_config, model, tokenizer):
+    assert OnnxGenAIChatCompletion(ai_model_path="/valid_path", template="phi3")
+
+
+def test_onnx_chat_completion_with_invalid_model():
+    with pytest.raises(ServiceInitializationError):
+        OnnxGenAIChatCompletion(
+            ai_model_path="/invalid_path",
+            template=ONNXTemplate.PHI3,
+        )
+
+
+@patch("builtins.open", new_callable=mock_open, read_data=json.dumps(gen_ai_config_vision))
+def test_onnx_chat_completion_with_multimodality_without_prompt_template(gen_ai_config_vision):
+    with pytest.raises(ServiceInitializationError):
+        OnnxGenAIChatCompletion()
+
+
+def test_onnx_chat_completion_with_invalid_env_variable(onnx_unit_test_env):
+    with pytest.raises(ServiceInitializationError):
+        OnnxGenAIChatCompletion(
+            template=ONNXTemplate.PHI3,
+        )
+
+
+@pytest.mark.parametrize("exclude_list", [["ONNX_GEN_AI_CHAT_MODEL_FOLDER"]], indirect=True)
+def test_onnx_chat_completion_with_missing_ai_path(onnx_unit_test_env):
+    with pytest.raises(ServiceInitializationError):
+        OnnxGenAIChatCompletion(template=ONNXTemplate.PHI3, env_file_path="test.env")
+
+
+@patch("builtins.open", new_callable=mock_open, read_data=json.dumps(gen_ai_config))
+@patch("onnxruntime_genai.Model")
+@patch("onnxruntime_genai.Tokenizer")
+async def test_onnx_chat_completion(gen_ai_config, model, tokenizer):
+    generator_mock = MagicMock()
+    generator_mock.__aiter__.return_value = [["H"], ["e"], ["l"], ["l"], ["o"]]
+
+    chat_completion = OnnxGenAIChatCompletion(template=ONNXTemplate.PHI3, ai_model_path="test")
+
+    history = ChatHistory()
+    history.add_system_message("test")
+    history.add_user_message("test")
+
+    with patch.object(chat_completion, "_generate_next_token_async", return_value=generator_mock):
+        completed_text: ChatMessageContent = await chat_completion.get_chat_message_content(
+            prompt="test", chat_history=history, settings=OnnxGenAIPromptExecutionSettings(), kernel=Kernel()
+        )
+
+    assert str(completed_text) == "Hello"
+
+
+@patch("builtins.open", new_callable=mock_open, read_data=json.dumps(gen_ai_config))
+@patch("onnxruntime_genai.Model")
+@patch("onnxruntime_genai.Tokenizer")
+async def test_onnx_chat_completion_streaming(gen_ai_config, model, tokenizer):
+    generator_mock = MagicMock()
+    generator_mock.__aiter__.return_value = [["H"], ["e"], ["l"], ["l"], ["o"]]
+
+    chat_completion = OnnxGenAIChatCompletion(template=ONNXTemplate.PHI3, ai_model_path="test")
+
+    history = ChatHistory()
+    history.add_system_message("test")
+    history.add_user_message("test")
+
+    completed_text: str = ""
+
+    with patch.object(chat_completion, "_generate_next_token_async", return_value=generator_mock):
+        async for chunk in chat_completion.get_streaming_chat_message_content(
+            prompt="test", chat_history=history, settings=OnnxGenAIPromptExecutionSettings(), kernel=Kernel()
+        ):
+            completed_text += str(chunk)
+
+    assert completed_text == "Hello"
+
+
+@patch("onnxruntime_genai.Model")
+def test_onnx_chat_get_image_history(model):
+    builtin_open = open  # save the unpatched version
+
+    def patch_open(*args, **kwargs):
+        if "genai_config.json" in str(args[0]):
+            # mocked open for path "genai_config.json"
+            return mock_open(read_data=json.dumps(gen_ai_config_vision))(*args, **kwargs)
+        # unpatched version for every other path
+        return builtin_open(*args, **kwargs)
+
+    with patch("builtins.open", patch_open):
+        chat_completion = OnnxGenAIChatCompletion(
+            template=ONNXTemplate.PHI3,
+            ai_model_path="test",
+        )
+
+        image_content = ImageContent.from_image_path(
+            image_path=os.path.join(os.path.dirname(__file__), "../../../../../", "assets/sample_image.jpg")
+        )
+
+        history = ChatHistory()
+        history.add_system_message("test")
+        history.add_user_message("test")
+        history.add_message(
+            ChatMessageContent(
+                role=AuthorRole.USER,
+                items=[image_content],
+            ),
+        )
+
+        last_image = chat_completion._get_images_from_history(history)
+        assert last_image == [image_content]
+
+
+@patch("onnxruntime_genai.Model")
+@patch("onnxruntime_genai.Tokenizer")
+async def test_onnx_chat_get_image_history_with_not_multimodal(model, tokenizer):
+    builtin_open = open  # save the unpatched version
+
+    def patch_open(*args, **kwargs):
+        if "genai_config.json" in str(args[0]):
+            # mocked open for path "genai_config.json"
+            return mock_open(read_data=json.dumps(gen_ai_config))(*args, **kwargs)
+        # unpatched version for every other path
+        return builtin_open(*args, **kwargs)
+
+    with patch("builtins.open", patch_open):
+        chat_completion = OnnxGenAIChatCompletion(
+            template=ONNXTemplate.PHI3,
+            ai_model_path="test",
+        )
+
+        image_content = ImageContent.from_image_path(
+            image_path=os.path.join(os.path.dirname(__file__), "../../../../../", "assets/sample_image.jpg")
+        )
+
+        history = ChatHistory()
+        history.add_system_message("test")
+        history.add_user_message("test")
+        history.add_message(
+            ChatMessageContent(
+                role=AuthorRole.USER,
+                items=[image_content],
+            ),
+        )
+
+        with pytest.raises(ServiceInvalidExecutionSettingsError):
+            _ = await chat_completion._get_images_from_history(history)

python/tests/unit/connectors/ai/onnx/services/test_onnx_text_completion.py

@@ -0,0 +1,74 @@
+# Copyright (c) Microsoft. All rights reserved.
+import json
+from unittest.mock import MagicMock, mock_open, patch
+
+import pytest
+
+from semantic_kernel.connectors.ai.onnx import OnnxGenAIPromptExecutionSettings, OnnxGenAITextCompletion  # noqa: E402
+from semantic_kernel.contents import TextContent
+from semantic_kernel.exceptions import ServiceInitializationError
+from tests.unit.connectors.ai.onnx.conftest import gen_ai_config
+
+
+@patch("builtins.open", new_callable=mock_open, read_data=json.dumps(gen_ai_config))
+@patch("onnxruntime_genai.Model")
+@patch("onnxruntime_genai.Tokenizer")
+def test_onnx_chat_completion_with_valid_env_variable(gen_ai_config, model, tokenizer, onnx_unit_test_env):
+    assert OnnxGenAITextCompletion(env_file_path="test.env")
+
+
+@patch("builtins.open", new_callable=mock_open, read_data=json.dumps(gen_ai_config))
+@patch("onnxruntime_genai.Model")
+@patch("onnxruntime_genai.Tokenizer")
+def test_onnx_chat_completion_with_valid_parameter(gen_ai_config, model, tokenizer):
+    assert OnnxGenAITextCompletion(ai_model_path="/valid_path")
+
+
+def test_onnx_chat_completion_with_invalid_model():
+    with pytest.raises(ServiceInitializationError):
+        OnnxGenAITextCompletion(ai_model_path="/invalid_path")
+
+
+def test_onnx_chat_completion_with_invalid_env_variable(onnx_unit_test_env):
+    with pytest.raises(ServiceInitializationError):
+        OnnxGenAITextCompletion()
+
+
+@pytest.mark.parametrize("exclude_list", [["ONNX_GEN_AI_TEXT_MODEL_FOLDER"]], indirect=True)
+def test_onnx_chat_completion_with_missing_ai_path(onnx_unit_test_env):
+    with pytest.raises(ServiceInitializationError):
+        OnnxGenAITextCompletion(env_file_path="test.env")
+
+
+@patch("builtins.open", new_callable=mock_open, read_data=json.dumps(gen_ai_config))
+@patch("onnxruntime_genai.Model")
+@patch("onnxruntime_genai.Tokenizer")
+async def test_onnx_text_completion(gen_ai_config, model, tokenizer):
+    generator_mock = MagicMock()
+    generator_mock.__aiter__.return_value = [["H"], ["e"], ["l"], ["l"], ["o"]]
+
+    text_completion = OnnxGenAITextCompletion(ai_model_path="test")
+    with patch.object(text_completion, "_generate_next_token_async", return_value=generator_mock):
+        completed_text: TextContent = await text_completion.get_text_content(
+            prompt="test", settings=OnnxGenAIPromptExecutionSettings()
+        )
+
+    assert completed_text.text == "Hello"
+
+
+@patch("builtins.open", new_callable=mock_open, read_data=json.dumps(gen_ai_config))
+@patch("onnxruntime_genai.Model")
+@patch("onnxruntime_genai.Tokenizer")
+async def test_onnx_text_completion_streaming(gen_ai_config, model, tokenizer):
+    generator_mock = MagicMock()
+    generator_mock.__aiter__.return_value = [["H"], ["e"], ["l"], ["l"], ["o"]]
+
+    text_completion = OnnxGenAITextCompletion(ai_model_path="test")
+    completed_text: str = ""
+    with patch.object(text_completion, "_generate_next_token_async", return_value=generator_mock):
+        async for chunk in text_completion.get_streaming_text_content(
+            prompt="test", settings=OnnxGenAIPromptExecutionSettings()
+        ):
+            completed_text += chunk.text
+
+    assert completed_text == "Hello"

python/tests/unit/connectors/ai/onnx/services/test_onnx_utils.py

@@ -0,0 +1,132 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+from semantic_kernel.connectors.ai.onnx.utils import (
+    gemma_template,
+    llama_template,
+    phi3_template,
+    phi3v_template,
+    phi4_template,
+    phi4mm_template,
+)
+from semantic_kernel.contents import AudioContent, AuthorRole, ChatHistory, ImageContent, TextContent
+
+
+def test_phi3v_template_with_text_and_image():
+    history = ChatHistory(
+        messages=[
+            {"role": AuthorRole.SYSTEM, "content": "System message"},
+            {
+                "role": AuthorRole.USER,
+                "items": [TextContent(text="User text message"), ImageContent(url="http://example.com/image.png")],
+            },
+            {"role": AuthorRole.ASSISTANT, "content": "Assistant message"},
+        ]
+    )
+
+    expected_output = (
+        "<|system|>\nSystem message<|end|>\n"
+        "<|user|>\nUser text message<|end|>\n"
+        "<|image_1|>\n"
+        "<|assistant|>\nAssistant message<|end|>\n"
+        "<|assistant|>\n"
+    )
+
+    assert phi3v_template(history) == expected_output
+
+
+def test_phi4mm_template_with_text_and_image():
+    history = ChatHistory(
+        messages=[
+            {"role": AuthorRole.SYSTEM, "content": "System message"},
+            {
+                "role": AuthorRole.USER,
+                "items": [
+                    TextContent(text="User text message"),
+                    ImageContent(url="http://example.com/image.png"),
+                    AudioContent(url="http://example.com/audio.mp3"),
+                ],
+            },
+            {"role": AuthorRole.ASSISTANT, "content": "Assistant message"},
+        ]
+    )
+
+    expected_output = (
+        "<|system|>\nSystem message<|end|>\n"
+        "<|user|>\nUser text message<|end|>\n"
+        "<|image_1|>\n"
+        "<|audio_1|>\n"
+        "<|assistant|>\nAssistant message<|end|>\n"
+        "<|assistant|>\n"
+    )
+
+    assert phi4mm_template(history) == expected_output
+
+
+def test_phi3_template_with_only_text():
+    history = ChatHistory(messages=[{"role": AuthorRole.USER, "items": [TextContent(text="User text message")]}])
+
+    expected_output = "<|user|>\nUser text message<|end|>\n<|assistant|>\n"
+
+    assert phi3_template(history) == expected_output
+
+
+def test_phi4_template_with_only_text():
+    history = ChatHistory(messages=[{"role": AuthorRole.USER, "items": [TextContent(text="User text message")]}])
+
+    expected_output = "<|user|>\nUser text message<|end|>\n<|assistant|>\n"
+
+    assert phi4_template(history) == expected_output
+
+
+def test_gemma_template_with_user_and_assistant_messages():
+    history = ChatHistory(
+        messages=[
+            {"role": AuthorRole.USER, "content": "User text message"},
+            {"role": AuthorRole.ASSISTANT, "content": "Assistant message"},
+        ]
+    )
+
+    expected_output = (
+        "<bos>"
+        "<start_of_turn>user\nUser text message<end_of_turn>\n"
+        "<start_of_turn>model\nAssistant message<end_of_turn>\n"
+        "<start_of_turn>model\n"
+    )
+
+    assert gemma_template(history) == expected_output
+
+
+def test_gemma_template_with_only_user_message():
+    history = ChatHistory(messages=[{"role": AuthorRole.USER, "content": "User text message"}])
+
+    expected_output = "<bos><start_of_turn>user\nUser text message<end_of_turn>\n<start_of_turn>model\n"
+
+    assert gemma_template(history) == expected_output
+
+
+def test_llama_template_with_user_and_assistant_messages():
+    history = ChatHistory(
+        messages=[
+            {"role": AuthorRole.USER, "content": "User text message"},
+            {"role": AuthorRole.ASSISTANT, "content": "Assistant message"},
+        ]
+    )
+
+    expected_output = (
+        "<|start_header_id|>user<|end_header_id|>\n\nUser text message<|eot_id|>"
+        "<|start_header_id|>assistant<|end_header_id|>\n\nAssistant message<|eot_id|>"
+        "<|start_header_id|>assistant<|end_header_id|>"
+    )
+
+    assert llama_template(history) == expected_output
+
+
+def test_llama_template_with_only_user_message():
+    history = ChatHistory(messages=[{"role": AuthorRole.USER, "content": "User text message"}])
+
+    expected_output = (
+        "<|start_header_id|>user<|end_header_id|>\n\nUser text message<|eot_id|>"
+        "<|start_header_id|>assistant<|end_header_id|>"
+    )
+
+    assert llama_template(history) == expected_output

python/tests/unit/connectors/ai/onnx/test_onnx_prompt_execution_settings.py

@@ -0,0 +1,84 @@
+# Copyright (c) Microsoft. All rights reserved.
+import pytest
+from pydantic import ValidationError
+
+from semantic_kernel.connectors.ai.onnx.onnx_gen_ai_prompt_execution_settings import (
+    OnnxGenAIPromptExecutionSettings,
+)
+from semantic_kernel.connectors.ai.prompt_execution_settings import PromptExecutionSettings
+
+
+def test_default_onnx_chat_prompt_execution_settings():
+    settings = OnnxGenAIPromptExecutionSettings()
+    assert settings.temperature is None
+    assert settings.top_p is None
+
+
+def test_custom_onnx_chat_prompt_execution_settings():
+    settings = OnnxGenAIPromptExecutionSettings(
+        temperature=0.5,
+        top_p=0.5,
+        max_length=128,
+    )
+    assert settings.temperature == 0.5
+    assert settings.top_p == 0.5
+    assert settings.max_length == 128
+
+
+def test_onnx_chat_prompt_execution_settings_from_default_completion_config():
+    settings = PromptExecutionSettings(service_id="test_service")
+    chat_settings = OnnxGenAIPromptExecutionSettings.from_prompt_execution_settings(settings)
+    assert chat_settings.service_id == "test_service"
+    assert chat_settings.temperature is None
+    assert chat_settings.top_p is None
+
+
+def test_onnx_chat_prompt_execution_settings_from_onnx_prompt_execution_settings():
+    chat_settings = OnnxGenAIPromptExecutionSettings(service_id="test_service", temperature=1.0)
+    new_settings = OnnxGenAIPromptExecutionSettings(service_id="test_2", temperature=0.0)
+    chat_settings.update_from_prompt_execution_settings(new_settings)
+    assert chat_settings.service_id == "test_2"
+    assert chat_settings.temperature == 0.0
+
+
+def test_onnx_chat_prompt_execution_settings_from_custom_completion_config():
+    settings = PromptExecutionSettings(
+        service_id="test_service",
+        extension_data={
+            "temperature": 0.5,
+            "top_p": 0.5,
+            "max_length": 128,
+        },
+    )
+    chat_settings = OnnxGenAIPromptExecutionSettings.from_prompt_execution_settings(settings)
+    assert chat_settings.temperature == 0.5
+    assert chat_settings.top_p == 0.5
+    assert chat_settings.max_length == 128
+
+
+def test_create_options():
+    settings = OnnxGenAIPromptExecutionSettings(
+        service_id="test_service",
+        extension_data={
+            "temperature": 0.5,
+            "top_p": 0.5,
+            "max_length": 128,
+        },
+    )
+    options = settings.prepare_settings_dict()
+    assert options["temperature"] == 0.5
+    assert options["top_p"] == 0.5
+    assert options["max_length"] == 128
+
+
+def test_create_options_with_wrong_parameter():
+    with pytest.raises(ValidationError):
+        OnnxGenAIPromptExecutionSettings(
+            service_id="test_service",
+            function_choice_behavior="auto",
+            extension_data={
+                "temperature": 10.0,
+                "top_p": 0.5,
+                "max_length": 128,
+            },
+        )