chore: import upstream snapshot with attribution

python/semantic_kernel/connectors/ai/nvidia/README.md

@@ -0,0 +1,66 @@
+# semantic_kernel.connectors.ai.nvidia
+
+This connector enables integration with NVIDIA NIM API for text embeddings and chat completion. It allows you to use NVIDIA's models within the Semantic Kernel framework.
+
+## Quick start
+
+### Initialize the kernel
+```python
+import semantic_kernel as sk
+kernel = sk.Kernel()
+```
+
+### Add NVIDIA text embedding service
+You can provide your API key directly or through environment variables
+```python
+from semantic_kernel.connectors.ai.nvidia import NvidiaTextEmbedding
+
+embedding_service = NvidiaTextEmbedding(
+ai_model_id="nvidia/nv-embedqa-e5-v5", # Default model if not specified
+api_key="your-nvidia-api-key", # Can also use NVIDIA_API_KEY env variable
+service_id="nvidia-embeddings" # Optional service identifier
+)
+```
+
+### Add the embedding service to the kernel
+```python
+kernel.add_service(embedding_service)
+```
+
+### Generate embeddings for text
+```python
+texts = ["Hello, world!", "Semantic Kernel is awesome"]
+embeddings = await kernel.get_service("nvidia-embeddings").generate_embeddings(texts)
+```
+
+### Add NVIDIA chat completion service
+```python
+from semantic_kernel.connectors.ai.nvidia import NvidiaChatCompletion
+
+chat_service = NvidiaChatCompletion(
+    ai_model_id="meta/llama-3.1-8b-instruct", # Default model if not specified
+    api_key="your-nvidia-api-key", # Can also use NVIDIA_API_KEY env variable
+    service_id="nvidia-chat" # Optional service identifier
+)
+kernel.add_service(chat_service)
+```
+
+### Basic chat completion
+```python
+response = await kernel.invoke_prompt("Hello, how are you?")
+```
+
+### Using with Chat Completion Agent
+```python
+from semantic_kernel.agents import ChatCompletionAgent
+from semantic_kernel.connectors.ai.nvidia import NvidiaChatCompletion
+
+agent = ChatCompletionAgent(
+    service=NvidiaChatCompletion(),
+    name="SK-Assistant",
+    instructions="You are a helpful assistant.",
+)
+response = await agent.get_response(messages="Write a haiku about Semantic Kernel.")
+print(response.content)
+```
+

python/semantic_kernel/connectors/ai/nvidia/__init__.py

@@ -0,0 +1,19 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+from semantic_kernel.connectors.ai.nvidia.prompt_execution_settings.nvidia_prompt_execution_settings import (
+    NvidiaChatPromptExecutionSettings,
+    NvidiaEmbeddingPromptExecutionSettings,
+    NvidiaPromptExecutionSettings,
+)
+from semantic_kernel.connectors.ai.nvidia.services.nvidia_chat_completion import NvidiaChatCompletion
+from semantic_kernel.connectors.ai.nvidia.services.nvidia_text_embedding import NvidiaTextEmbedding
+from semantic_kernel.connectors.ai.nvidia.settings.nvidia_settings import NvidiaSettings
+
+__all__ = [
+    "NvidiaChatCompletion",
+    "NvidiaChatPromptExecutionSettings",
+    "NvidiaEmbeddingPromptExecutionSettings",
+    "NvidiaPromptExecutionSettings",
+    "NvidiaSettings",
+    "NvidiaTextEmbedding",
+]

python/semantic_kernel/connectors/ai/nvidia/prompt_execution_settings/__init__.py

@@ -0,0 +1 @@
+# Copyright (c) Microsoft. All rights reserved.

python/semantic_kernel/connectors/ai/nvidia/prompt_execution_settings/nvidia_prompt_execution_settings.py

@@ -0,0 +1,73 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+from typing import Annotated, Any, Literal
+
+from pydantic import BaseModel, Field
+
+from semantic_kernel.connectors.ai.prompt_execution_settings import PromptExecutionSettings
+
+
+class NvidiaPromptExecutionSettings(PromptExecutionSettings):
+    """Settings for NVIDIA prompt execution."""
+
+    format: Literal["json"] | None = None
+    options: dict[str, Any] | None = None
+
+
+class NvidiaEmbeddingPromptExecutionSettings(NvidiaPromptExecutionSettings):
+    """Settings for NVIDIA embedding prompt execution."""
+
+    input: str | list[str] | None = None
+    ai_model_id: Annotated[str | None, Field(serialization_alias="model")] = None
+    encoding_format: Literal["float", "base64"] = "float"
+    truncate: Literal["NONE", "START", "END"] = "NONE"
+    input_type: Literal["passage", "query"] = "query"  # required param with default value query
+    user: str | None = None
+    extra_headers: dict | None = None
+    extra_body: dict | None = None
+    timeout: float | None = None
+    dimensions: Annotated[int | None, Field(gt=0)] = None
+
+    def prepare_settings_dict(self, **kwargs) -> dict[str, Any]:
+        """Override only for embeddings to exclude input_type and truncate."""
+        return self.model_dump(
+            exclude={"service_id", "extension_data", "structured_json_response", "input_type", "truncate"},
+            exclude_none=True,
+            by_alias=True,
+        )
+
+
+class NvidiaChatPromptExecutionSettings(NvidiaPromptExecutionSettings):
+    """Settings for NVIDIA chat prompt execution."""
+
+    messages: list[dict[str, str]] | None = None
+    ai_model_id: Annotated[str | None, Field(serialization_alias="model")] = None
+    temperature: float | None = None
+    top_p: float | None = None
+    n: int | None = None
+    stream: bool = False
+    stop: str | list[str] | None = None
+    max_tokens: int | None = None
+    presence_penalty: float | None = None
+    frequency_penalty: float | None = None
+    logit_bias: dict[str, float] | None = None
+    user: str | None = None
+    tools: list[dict[str, Any]] | None = None
+    tool_choice: str | dict[str, Any] | None = None
+    response_format: (
+        dict[Literal["type"], Literal["text", "json_object"]] | dict[str, Any] | type[BaseModel] | type | None
+    ) = None
+    seed: int | None = None
+    extra_headers: dict | None = None
+    extra_body: dict | None = None
+    timeout: float | None = None
+    # NVIDIA-specific structured output support
+    nvext: dict[str, Any] | None = None
+
+    def prepare_settings_dict(self, **kwargs) -> dict[str, Any]:
+        """Override only for embeddings to exclude input_type and truncate."""
+        return self.model_dump(
+            exclude={"service_id", "extension_data", "structured_json_response", "response_format"},
+            exclude_none=True,
+            by_alias=True,
+        )

python/semantic_kernel/connectors/ai/nvidia/services/__init__.py

@@ -0,0 +1 @@
+# Copyright (c) Microsoft. All rights reserved.

python/semantic_kernel/connectors/ai/nvidia/services/nvidia_chat_completion.py

@@ -0,0 +1,313 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+import logging
+import sys
+from collections.abc import AsyncGenerator
+from typing import Any, Literal
+
+from openai import AsyncOpenAI
+from openai.types.chat.chat_completion import ChatCompletion, Choice
+from openai.types.chat.chat_completion_chunk import ChatCompletionChunk
+from openai.types.chat.chat_completion_chunk import Choice as ChunkChoice
+from pydantic import ValidationError
+
+from semantic_kernel.connectors.ai.chat_completion_client_base import ChatCompletionClientBase
+from semantic_kernel.connectors.ai.completion_usage import CompletionUsage
+from semantic_kernel.connectors.ai.nvidia.prompt_execution_settings.nvidia_prompt_execution_settings import (
+    NvidiaChatPromptExecutionSettings,
+)
+from semantic_kernel.connectors.ai.nvidia.services.nvidia_handler import NvidiaHandler
+from semantic_kernel.connectors.ai.nvidia.services.nvidia_model_types import NvidiaModelTypes
+from semantic_kernel.connectors.ai.nvidia.settings.nvidia_settings import NvidiaSettings
+from semantic_kernel.connectors.ai.prompt_execution_settings import PromptExecutionSettings
+from semantic_kernel.contents import (
+    AuthorRole,
+    ChatMessageContent,
+    FinishReason,
+    FunctionCallContent,
+    StreamingChatMessageContent,
+    StreamingTextContent,
+    TextContent,
+)
+from semantic_kernel.contents.chat_history import ChatHistory
+from semantic_kernel.exceptions.service_exceptions import ServiceInitializationError
+from semantic_kernel.utils.feature_stage_decorator import experimental
+from semantic_kernel.utils.telemetry.model_diagnostics.decorators import (
+    trace_chat_completion,
+    trace_streaming_chat_completion,
+)
+
+if sys.version_info >= (3, 12):
+    from typing import override  # pragma: no cover
+else:
+    from typing_extensions import override  # pragma: no cover
+
+logger: logging.Logger = logging.getLogger(__name__)
+
+# Default NVIDIA chat model when none is specified
+DEFAULT_NVIDIA_CHAT_MODEL = "meta/llama-3.1-8b-instruct"
+
+
+@experimental
+class NvidiaChatCompletion(NvidiaHandler, ChatCompletionClientBase):
+    """NVIDIA Chat completion class.
+
+    This class does not support function calling. The SUPPORTS_FUNCTION_CALLING attribute
+    is set to False (inherited from the base class).
+    """
+
+    def __init__(
+        self,
+        ai_model_id: str | None = None,
+        api_key: str | None = None,
+        base_url: str | None = None,
+        service_id: str | None = None,
+        client: AsyncOpenAI | None = None,
+        env_file_path: str | None = None,
+        env_file_encoding: str | None = None,
+        instruction_role: Literal["system", "user", "assistant", "developer"] | None = None,
+    ) -> None:
+        """Initialize an NvidiaChatCompletion service.
+
+        Args:
+            ai_model_id (str): NVIDIA model name, see
+                https://docs.api.nvidia.com/nim/reference/
+                If not provided, defaults to DEFAULT_NVIDIA_CHAT_MODEL.
+            service_id (str | None): Service ID tied to the execution settings.
+            api_key (str | None): The optional API key to use. If provided will override,
+                the env vars or .env file value.
+            base_url (str | None): Custom API endpoint. (Optional)
+            client (Optional[AsyncOpenAI]): An existing client to use. (Optional)
+            env_file_path (str | None): Use the environment settings file as a fallback
+                to environment variables. (Optional)
+            env_file_encoding (str | None): The encoding of the environment settings file. (Optional)
+            instruction_role (Literal["system", "user", "assistant", "developer"] | None): The role to use for
+                'instruction' messages. Defaults to "system". (Optional)
+        """
+        try:
+            nvidia_settings = NvidiaSettings(
+                api_key=api_key,
+                base_url=base_url,
+                chat_model_id=ai_model_id,
+                env_file_path=env_file_path,
+                env_file_encoding=env_file_encoding,
+            )
+        except ValidationError as ex:
+            raise ServiceInitializationError("Failed to create NVIDIA settings.", ex) from ex
+
+        if not client and not nvidia_settings.api_key:
+            raise ServiceInitializationError("The NVIDIA API key is required.")
+        if not nvidia_settings.chat_model_id:
+            # Default fallback model
+            nvidia_settings.chat_model_id = DEFAULT_NVIDIA_CHAT_MODEL
+            logger.warning(f"Default chat model set as: {nvidia_settings.chat_model_id}")
+
+        # Create client if not provided
+        if not client:
+            client = AsyncOpenAI(
+                api_key=nvidia_settings.api_key.get_secret_value() if nvidia_settings.api_key else None,
+                base_url=nvidia_settings.base_url,
+            )
+
+        super().__init__(
+            ai_model_id=nvidia_settings.chat_model_id,
+            api_key=nvidia_settings.api_key.get_secret_value() if nvidia_settings.api_key else None,
+            base_url=nvidia_settings.base_url,
+            service_id=service_id or "",
+            ai_model_type=NvidiaModelTypes.CHAT,
+            client=client,
+            instruction_role=instruction_role or "system",
+        )
+
+    @classmethod
+    def from_dict(cls: type["NvidiaChatCompletion"], settings: dict[str, Any]) -> "NvidiaChatCompletion":
+        """Initialize an NVIDIA service from a dictionary of settings.
+
+        Args:
+            settings: A dictionary of settings for the service.
+        """
+        return cls(
+            ai_model_id=settings.get("ai_model_id"),
+            api_key=settings.get("api_key"),
+            base_url=settings.get("base_url"),
+            service_id=settings.get("service_id"),
+            env_file_path=settings.get("env_file_path"),
+        )
+
+    @override
+    def get_prompt_execution_settings_class(self) -> type["PromptExecutionSettings"]:
+        return NvidiaChatPromptExecutionSettings
+
+    @override
+    @trace_chat_completion("nvidia")
+    async def _inner_get_chat_message_contents(
+        self,
+        chat_history: "ChatHistory",
+        settings: "PromptExecutionSettings",
+    ) -> list["ChatMessageContent"]:
+        if not isinstance(settings, NvidiaChatPromptExecutionSettings):
+            settings = self.get_prompt_execution_settings_from_settings(settings)
+        assert isinstance(settings, NvidiaChatPromptExecutionSettings)  # nosec
+
+        settings.stream = False
+        settings.messages = self._prepare_chat_history_for_request(chat_history)
+        settings.ai_model_id = settings.ai_model_id or self.ai_model_id
+
+        # Handle structured output
+        self._handle_structured_output(settings)
+
+        response = await self._send_request(settings)
+        assert isinstance(response, ChatCompletion)  # nosec
+        response_metadata = self._get_metadata_from_chat_response(response)
+        return [self._create_chat_message_content(response, choice, response_metadata) for choice in response.choices]
+
+    @override
+    @trace_streaming_chat_completion("nvidia")
+    async def _inner_get_streaming_chat_message_contents(
+        self,
+        chat_history: "ChatHistory",
+        settings: "PromptExecutionSettings",
+        function_invoke_attempt: int = 0,
+    ) -> AsyncGenerator[list["StreamingChatMessageContent"], Any]:
+        if not isinstance(settings, NvidiaChatPromptExecutionSettings):
+            settings = self.get_prompt_execution_settings_from_settings(settings)
+        assert isinstance(settings, NvidiaChatPromptExecutionSettings)  # nosec
+
+        settings.stream = True
+        settings.messages = self._prepare_chat_history_for_request(chat_history)
+        settings.ai_model_id = settings.ai_model_id or self.ai_model_id
+
+        # Handle structured output
+        self._handle_structured_output(settings)
+
+        response = await self._send_request(settings)
+        assert isinstance(response, AsyncGenerator)  # nosec
+
+        async for chunk in response:
+            if len(chunk.choices) == 0:
+                continue
+            chunk_metadata = self._get_metadata_from_chat_response(chunk)
+            yield [
+                self._create_streaming_chat_message_content(chunk, choice, chunk_metadata, function_invoke_attempt)
+                for choice in chunk.choices
+            ]
+
+    def _create_chat_message_content(
+        self, response: ChatCompletion, choice: Choice, response_metadata: dict[str, Any]
+    ) -> "ChatMessageContent":
+        """Create a chat message content object from a choice."""
+        metadata = self._get_metadata_from_chat_choice(choice)
+        metadata.update(response_metadata)
+
+        items: list[Any] = self._get_tool_calls_from_chat_choice(choice)
+        items.extend(self._get_function_call_from_chat_choice(choice))
+        if choice.message.content:
+            items.append(TextContent(text=choice.message.content))
+
+        return ChatMessageContent(
+            inner_content=response,
+            ai_model_id=self.ai_model_id,
+            metadata=metadata,
+            role=AuthorRole(choice.message.role),
+            items=items,
+            finish_reason=(FinishReason(choice.finish_reason) if choice.finish_reason else None),
+        )
+
+    def _create_streaming_chat_message_content(
+        self,
+        chunk: ChatCompletionChunk,
+        choice: ChunkChoice,
+        chunk_metadata: dict[str, Any],
+        function_invoke_attempt: int,
+    ) -> StreamingChatMessageContent:
+        """Create a streaming chat message content object from a choice."""
+        metadata = self._get_metadata_from_chat_choice(choice)
+        metadata.update(chunk_metadata)
+
+        items: list[Any] = self._get_tool_calls_from_chat_choice(choice)
+        items.extend(self._get_function_call_from_chat_choice(choice))
+        if choice.delta and choice.delta.content is not None:
+            items.append(StreamingTextContent(choice_index=choice.index, text=choice.delta.content))
+        return StreamingChatMessageContent(
+            choice_index=choice.index,
+            inner_content=chunk,
+            ai_model_id=self.ai_model_id,
+            metadata=metadata,
+            role=(AuthorRole(choice.delta.role) if choice.delta and choice.delta.role else AuthorRole.ASSISTANT),
+            finish_reason=(FinishReason(choice.finish_reason) if choice.finish_reason else None),
+            items=items,
+            function_invoke_attempt=function_invoke_attempt,
+        )
+
+    def _get_metadata_from_chat_response(self, response: ChatCompletion | ChatCompletionChunk) -> dict[str, Any]:
+        """Get metadata from a chat response."""
+        return {
+            "id": response.id,
+            "created": response.created,
+            "system_fingerprint": getattr(response, "system_fingerprint", None),
+            "usage": CompletionUsage.from_openai(response.usage) if response.usage is not None else None,
+        }
+
+    def _get_metadata_from_chat_choice(self, choice: Choice | ChunkChoice) -> dict[str, Any]:
+        """Get metadata from a chat choice."""
+        return {
+            "logprobs": getattr(choice, "logprobs", None),
+        }
+
+    def _get_tool_calls_from_chat_choice(self, choice: Choice | ChunkChoice) -> list[FunctionCallContent]:
+        """Get tool calls from a chat choice."""
+        content = choice.message if isinstance(choice, Choice) else choice.delta
+        if content and (tool_calls := getattr(content, "tool_calls", None)) is not None:
+            return [
+                FunctionCallContent(
+                    id=tool.id,
+                    index=getattr(tool, "index", None),
+                    name=tool.function.name,
+                    arguments=tool.function.arguments,
+                )
+                for tool in tool_calls
+            ]
+        return []
+
+    def _get_function_call_from_chat_choice(self, choice: Choice | ChunkChoice) -> list[FunctionCallContent]:
+        """Get function calls from a chat choice."""
+        content = choice.message if isinstance(choice, Choice) else choice.delta
+        if content and (function_call := getattr(content, "function_call", None)) is not None:
+            return [
+                FunctionCallContent(
+                    id="",
+                    name=function_call.name,
+                    arguments=function_call.arguments,
+                )
+            ]
+        return []
+
+    def _handle_structured_output(self, request_settings: NvidiaChatPromptExecutionSettings) -> None:
+        """Handle structured output for NVIDIA models using nvext parameter."""
+        response_format = getattr(request_settings, "response_format", None)
+        if response_format:
+            # Convert Pydantic model to JSON schema for NVIDIA's guided_json
+            if hasattr(response_format, "model_json_schema"):
+                # It's a Pydantic model
+                schema = response_format.model_json_schema()
+                if not request_settings.extra_body:
+                    request_settings.extra_body = {}
+                request_settings.extra_body["nvext"] = {"guided_json": schema}
+            elif isinstance(response_format, dict):
+                # It's already a dict, use it directly
+                if not request_settings.extra_body:
+                    request_settings.extra_body = {}
+                request_settings.extra_body["nvext"] = {"guided_json": response_format}
+
+    def _prepare_chat_history_for_request(
+        self,
+        chat_history: ChatHistory,
+        role_key: str = "role",
+        content_key: str = "content",
+    ) -> list[dict[str, str]]:
+        """Prepare chat history for request."""
+        messages = []
+        for message in chat_history.messages:
+            message_dict = {role_key: message.role.value, content_key: message.content}
+            messages.append(message_dict)
+        return messages

python/semantic_kernel/connectors/ai/nvidia/services/nvidia_handler.py

@@ -0,0 +1,123 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+import logging
+from abc import ABC
+from typing import Any, ClassVar, Union
+
+from openai import AsyncOpenAI, AsyncStream
+from openai.types.chat.chat_completion import ChatCompletion
+from openai.types.chat.chat_completion_chunk import ChatCompletionChunk
+from openai.types.completion import Completion
+from openai.types.create_embedding_response import CreateEmbeddingResponse
+
+from semantic_kernel.connectors.ai.nvidia.prompt_execution_settings.nvidia_prompt_execution_settings import (
+    NvidiaChatPromptExecutionSettings,
+    NvidiaEmbeddingPromptExecutionSettings,
+)
+from semantic_kernel.connectors.ai.nvidia.services.nvidia_model_types import NvidiaModelTypes
+from semantic_kernel.connectors.ai.prompt_execution_settings import PromptExecutionSettings
+from semantic_kernel.const import USER_AGENT
+from semantic_kernel.exceptions import ServiceResponseException
+from semantic_kernel.kernel_pydantic import KernelBaseModel
+
+logger: logging.Logger = logging.getLogger(__name__)
+
+RESPONSE_TYPE = Union[list[Any], ChatCompletion, Completion, AsyncStream[Any]]
+
+
+class NvidiaHandler(KernelBaseModel, ABC):
+    """Internal class for calls to Nvidia API's."""
+
+    MODEL_PROVIDER_NAME: ClassVar[str] = "nvidia"
+    client: AsyncOpenAI
+    ai_model_type: NvidiaModelTypes = NvidiaModelTypes.CHAT
+    completion_tokens: int = 0
+    total_tokens: int = 0
+    prompt_tokens: int = 0
+
+    async def _send_request(self, settings: PromptExecutionSettings) -> RESPONSE_TYPE:
+        """Send a request to the Nvidia API."""
+        if self.ai_model_type == NvidiaModelTypes.EMBEDDING:
+            assert isinstance(settings, NvidiaEmbeddingPromptExecutionSettings)  # nosec
+            return await self._send_embedding_request(settings)
+        if self.ai_model_type == NvidiaModelTypes.CHAT:
+            assert isinstance(settings, NvidiaChatPromptExecutionSettings)  # nosec
+            return await self._send_chat_completion_request(settings)
+
+        raise NotImplementedError(f"Model type {self.ai_model_type} is not supported")
+
+    async def _send_embedding_request(self, settings: NvidiaEmbeddingPromptExecutionSettings) -> list[Any]:
+        """Send a request to the OpenAI embeddings endpoint."""
+        try:
+            # unsupported parameters are internally excluded from main dict and added to extra_body
+            response = await self.client.embeddings.create(**settings.prepare_settings_dict())
+
+            self.store_usage(response)
+            return [x.embedding for x in response.data]
+        except Exception as ex:
+            raise ServiceResponseException(
+                f"{type(self)} service failed to generate embeddings",
+                ex,
+            ) from ex
+
+    async def _send_chat_completion_request(
+        self, settings: NvidiaChatPromptExecutionSettings
+    ) -> ChatCompletion | AsyncStream[Any]:
+        """Send a request to the NVIDIA chat completion endpoint."""
+        try:
+            settings_dict = settings.prepare_settings_dict()
+
+            # Handle structured output if nvext is present in extra_body
+            if settings.extra_body and "nvext" in settings.extra_body:
+                if "extra_body" not in settings_dict:
+                    settings_dict["extra_body"] = {}
+                settings_dict["extra_body"]["nvext"] = settings.extra_body["nvext"]
+
+            response = await self.client.chat.completions.create(**settings_dict)
+            self.store_usage(response)
+            return response
+        except Exception as ex:
+            raise ServiceResponseException(
+                f"{type(self)} service failed to complete the chat",
+                ex,
+            ) from ex
+
+    def store_usage(
+        self,
+        response: ChatCompletion
+        | Completion
+        | AsyncStream[ChatCompletionChunk]
+        | AsyncStream[Completion]
+        | CreateEmbeddingResponse,
+    ):
+        """Store the usage information from the response."""
+        if not isinstance(response, AsyncStream) and response.usage:
+            logger.info(f"OpenAI usage: {response.usage}")
+            self.prompt_tokens += response.usage.prompt_tokens
+            self.total_tokens += response.usage.total_tokens
+            if hasattr(response.usage, "completion_tokens"):
+                self.completion_tokens += response.usage.completion_tokens
+
+    def to_dict(self) -> dict[str, str]:
+        """Create a dict of the service settings."""
+        client_settings = {
+            "api_key": self.client.api_key,
+            "default_headers": {k: v for k, v in self.client.default_headers.items() if k != USER_AGENT},
+        }
+        if self.client.organization:
+            client_settings["org_id"] = self.client.organization
+        base = self.model_dump(
+            exclude={
+                "prompt_tokens",
+                "completion_tokens",
+                "total_tokens",
+                "api_type",
+                "ai_model_type",
+                "service_id",
+                "client",
+            },
+            by_alias=True,
+            exclude_none=True,
+        )
+        base.update(client_settings)
+        return base

python/semantic_kernel/connectors/ai/nvidia/services/nvidia_model_types.py

@@ -0,0 +1,10 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+from enum import Enum
+
+
+class NvidiaModelTypes(Enum):
+    """Nvidia model types, can be text, chat, or embedding."""
+
+    EMBEDDING = "embedding"
+    CHAT = "chat"

python/semantic_kernel/connectors/ai/nvidia/services/nvidia_text_embedding.py

@@ -0,0 +1,167 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+import asyncio
+import copy
+import logging
+import sys
+from typing import Any
+
+if sys.version_info >= (3, 12):
+    from typing import override  # pragma: no cover
+else:
+    from typing_extensions import override  # pragma: no cover
+
+from numpy import array, ndarray
+from openai import AsyncOpenAI
+from pydantic import ValidationError
+
+from semantic_kernel.connectors.ai.embedding_generator_base import EmbeddingGeneratorBase
+from semantic_kernel.connectors.ai.nvidia.prompt_execution_settings.nvidia_prompt_execution_settings import (
+    NvidiaEmbeddingPromptExecutionSettings,
+)
+from semantic_kernel.connectors.ai.nvidia.services.nvidia_handler import NvidiaHandler
+from semantic_kernel.connectors.ai.nvidia.services.nvidia_model_types import NvidiaModelTypes
+from semantic_kernel.connectors.ai.nvidia.settings.nvidia_settings import NvidiaSettings
+from semantic_kernel.connectors.ai.prompt_execution_settings import PromptExecutionSettings
+from semantic_kernel.exceptions.service_exceptions import ServiceInitializationError
+from semantic_kernel.utils.feature_stage_decorator import experimental
+
+logger: logging.Logger = logging.getLogger(__name__)
+
+
+@experimental
+class NvidiaTextEmbedding(NvidiaHandler, EmbeddingGeneratorBase):
+    """Nvidia text embedding service."""
+
+    def __init__(
+        self,
+        ai_model_id: str | None = None,
+        api_key: str | None = None,
+        base_url: str | None = None,
+        client: AsyncOpenAI | None = None,
+        env_file_path: str | None = None,
+        service_id: str | None = None,
+    ) -> None:
+        """Initializes a new instance of the NvidiaTextEmbedding class.
+
+        Args:
+            ai_model_id (str): NVIDIA model card string, see
+                https://Nvidia.co/sentence-transformers
+            api_key: NVIDIA API key, see https://console.NVIDIA.com/settings/keys
+                (Env var NVIDIA_API_KEY)
+            base_url: HttpsUrl | None - base_url: The url of the NVIDIA endpoint. The base_url consists of the endpoint,
+                and more information refer https://docs.api.nvidia.com/nim/reference/
+                use endpoint if you only want to supply the endpoint.
+                (Env var NVIDIA_BASE_URL)
+            client (Optional[AsyncOpenAI]): An existing client to use. (Optional)
+            env_file_path (str | None): Use the environment settings file as
+                a fallback to environment variables. (Optional)
+            service_id (str): Service ID for the model. (optional)
+        """
+        try:
+            nvidia_settings = NvidiaSettings(
+                api_key=api_key,
+                base_url=base_url,
+                embedding_model_id=ai_model_id,
+                env_file_path=env_file_path,
+            )
+        except ValidationError as ex:
+            raise ServiceInitializationError("Failed to create NVIDIA settings.", ex) from ex
+        if not nvidia_settings.embedding_model_id:
+            nvidia_settings.embedding_model_id = "nvidia/nv-embedqa-e5-v5"
+            logger.warning(f"Default embedding model set as: {nvidia_settings.embedding_model_id}")
+        if not nvidia_settings.api_key:
+            logger.warning("API_KEY is missing, inference may fail.")
+        if not client:
+            client = AsyncOpenAI(
+                api_key=nvidia_settings.api_key.get_secret_value() if nvidia_settings.api_key else None,
+                base_url=nvidia_settings.base_url,
+            )
+        super().__init__(
+            ai_model_id=nvidia_settings.embedding_model_id,
+            api_key=nvidia_settings.api_key.get_secret_value() if nvidia_settings.api_key else None,
+            ai_model_type=NvidiaModelTypes.EMBEDDING,
+            service_id=service_id or nvidia_settings.embedding_model_id,
+            env_file_path=env_file_path,
+            client=client,
+        )
+
+    @override
+    async def generate_embeddings(
+        self,
+        texts: list[str],
+        settings: "PromptExecutionSettings | None" = None,
+        batch_size: int | None = None,
+        **kwargs: Any,
+    ) -> ndarray:
+        raw_embeddings = await self.generate_raw_embeddings(texts, settings, batch_size, **kwargs)
+        return array(raw_embeddings)
+
+    @override
+    async def generate_raw_embeddings(
+        self,
+        texts: list[str],
+        settings: "PromptExecutionSettings | None" = None,
+        batch_size: int | None = None,
+        **kwargs: Any,
+    ) -> Any:
+        """Returns embeddings for the given texts in the unedited format.
+
+        Args:
+            texts (List[str]): The texts to generate embeddings for.
+            settings (NvidiaEmbeddingPromptExecutionSettings): The settings to use for the request.
+            batch_size (int): The batch size to use for the request.
+            kwargs (Dict[str, Any]): Additional arguments to pass to the request.
+        """
+        if not settings:
+            settings = NvidiaEmbeddingPromptExecutionSettings(ai_model_id=self.ai_model_id)
+        else:
+            if not isinstance(settings, NvidiaEmbeddingPromptExecutionSettings):
+                settings = self.get_prompt_execution_settings_from_settings(settings)
+        assert isinstance(settings, NvidiaEmbeddingPromptExecutionSettings)  # nosec
+        if settings.ai_model_id is None:
+            settings.ai_model_id = self.ai_model_id
+        for key, value in kwargs.items():
+            setattr(settings, key, value)
+
+        # move input_type and truncate to extra-body
+        if not settings.extra_body:
+            settings.extra_body = {}
+        settings.extra_body.setdefault("input_type", settings.input_type)
+        if settings.truncate is not None:
+            settings.extra_body.setdefault("truncate", settings.truncate)
+
+        raw_embeddings = []
+        tasks = []
+
+        batch_size = batch_size or len(texts)
+        for i in range(0, len(texts), batch_size):
+            batch = texts[i : i + batch_size]
+            batch_settings = copy.deepcopy(settings)
+            batch_settings.input = batch
+            tasks.append(self._send_request(settings=batch_settings))
+
+        results = await asyncio.gather(*tasks)
+        for raw_embedding in results:
+            assert isinstance(raw_embedding, list)  # nosec
+            raw_embeddings.extend(raw_embedding)
+
+        return raw_embeddings
+
+    def get_prompt_execution_settings_class(self) -> type["PromptExecutionSettings"]:
+        """Get the request settings class."""
+        return NvidiaEmbeddingPromptExecutionSettings
+
+    @classmethod
+    def from_dict(cls: type["NvidiaTextEmbedding"], settings: dict[str, Any]) -> "NvidiaTextEmbedding":
+        """Initialize an Open AI service from a dictionary of settings.
+
+        Args:
+            settings: A dictionary of settings for the service.
+        """
+        return cls(
+            ai_model_id=settings.get("ai_model_id"),
+            api_key=settings.get("api_key"),
+            env_file_path=settings.get("env_file_path"),
+            service_id=settings.get("service_id"),
+        )

python/semantic_kernel/connectors/ai/nvidia/settings/__init__.py

@@ -0,0 +1 @@
+# Copyright (c) Microsoft. All rights reserved.

python/semantic_kernel/connectors/ai/nvidia/settings/nvidia_settings.py

@@ -0,0 +1,37 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+from typing import ClassVar
+
+from pydantic import SecretStr
+
+from semantic_kernel.kernel_pydantic import KernelBaseSettings
+
+
+class NvidiaSettings(KernelBaseSettings):
+    """Nvidia model settings.
+
+    The settings are first loaded from environment variables with the prefix 'NVIDIA_'. If the
+    environment variables are not found, the settings can be loaded from a .env file with the
+    encoding 'utf-8'. If the settings are not found in the .env file, the settings are ignored;
+    however, validation will fail alerting that the settings are missing.
+
+    Optional settings for prefix 'NVIDIA_' are:
+    - api_key: NVIDIA API key, see https://console.NVIDIA.com/settings/keys
+        (Env var NVIDIA_API_KEY)
+    - base_url: HttpsUrl | None - base_url: The url of the NVIDIA endpoint. The base_url consists of the endpoint,
+                and more information refer https://docs.api.nvidia.com/nim/reference/
+                use endpoint if you only want to supply the endpoint.
+                (Env var NVIDIA_BASE_URL)
+    - embedding_model_id: str | None - The NVIDIA embedding model ID to use, for example, nvidia/nv-embed-v1.
+        (Env var NVIDIA_EMBEDDING_MODEL_ID)
+    - chat_model_id: str | None - The NVIDIA chat model ID to use.
+        (Env var NVIDIA_CHAT_MODEL_ID)
+    - env_file_path: if provided, the .env settings are read from this file path location
+    """
+
+    env_prefix: ClassVar[str] = "NVIDIA_"
+
+    api_key: SecretStr | None = None
+    base_url: str = "https://integrate.api.nvidia.com/v1"
+    embedding_model_id: str | None = None
+    chat_model_id: str | None = None