Python: Introduce NvidiaChatCompletion AI Connector (#12952)

### 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?
Enables developers to use NVIDIA's chat models through Semantic Kernel
with structured output support.
  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.
-->
This change adds chat completion support with structured output to the
existing NVIDIA connector, enabling developers to use NVIDIA's chat
models through Semantic Kernel while building conversational AI
applications that require structured, validated responses. The connector
will support more model types iteratively, with future PRs introducing
VLM models and additional capabilities.

### Description

<!-- Describe your changes, the overall approach, the underlying design.
These notes will help understanding how your code works. Thanks! -->

- New Chat Completion Service:
`NvidiaChatCompletion`: Chat completion service following same structure
as other connectors
- Extends existing NVIDIA embedding connector patterns and architecture
- Custom logic for handling structured output specific to NVIDIA models
  
- Enhanced Configuration:
`NvidiaChatPromptExecutionSettings`: Chat-specific settings following
existing connector patterns
### 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 😄

Author: soumilinandi

python/samples/concepts/setup/ALL_SETTINGS.md

@@ -30,7 +30,8 @@
 |  | [VertexAITextEmbedding](../../../semantic_kernel/connectors/ai/google/google_ai/services/google_ai_text_embedding.py) | project_id, <br> region, <br> embedding_model_id | VERTEX_AI_PROJECT_ID, <br> VERTEX_AI_REGION, <br> VERTEX_AI_EMBEDDING_MODEL_ID | Yes, <br> No, <br> Yes |  |
 | HuggingFace | [HuggingFaceTextCompletion](../../../semantic_kernel/connectors/ai/hugging_face/services/hf_text_completion.py) | ai_model_id | N/A | Yes | |
 |  | [HuggingFaceTextEmbedding](../../../semantic_kernel/connectors/ai/hugging_face/services/hf_text_embedding.py) | ai_model_id | N/A | Yes | |
-| NVIDIA NIM | [NvidiaTextEmbedding](../../../semantic_kernel/connectors/ai/nvidia/services/nvidia_text_embedding.py) | ai_model_id, <br> api_key, <br> base_url | NVIDIA_API_KEY, <br> NVIDIA_TEXT_EMBEDDING_MODEL_ID, <br> NVIDIA_BASE_URL | Yes | [NvidiaAISettings](../../../semantic_kernel/connectors/ai/nvidia/settings/nvidia_settings.py) |
+| NVIDIA NIM | [NvidiaChatCompletion](../../../semantic_kernel/connectors/ai/nvidia/services/nvidia_chat_completion.py) | ai_model_id, <br> api_key, <br> base_url | NVIDIA_CHAT_MODEL_ID, <br> NVIDIA_API_KEY, <br> NVIDIA_BASE_URL | Yes (default: meta/llama-3.1-8b-instruct), <br> Yes, <br> No | [NvidiaAISettings](../../../semantic_kernel/connectors/ai/nvidia/settings/nvidia_settings.py) |
+|  | [NvidiaTextEmbedding](../../../semantic_kernel/connectors/ai/nvidia/services/nvidia_text_embedding.py) | ai_model_id, <br> api_key, <br> base_url | NVIDIA_API_KEY, <br> NVIDIA_TEXT_EMBEDDING_MODEL_ID, <br> NVIDIA_BASE_URL | Yes | [NvidiaAISettings](../../../semantic_kernel/connectors/ai/nvidia/settings/nvidia_settings.py) |
 | Mistral AI | [MistralAIChatCompletion](../../../semantic_kernel/connectors/ai/mistral_ai/services/mistral_ai_chat_completion.py) | ai_model_id, <br> api_key | MISTRALAI_CHAT_MODEL_ID, <br> MISTRALAI_API_KEY | Yes, <br> Yes | [MistralAISettings](../../../semantic_kernel/connectors/ai/mistral_ai/settings/mistral_ai_settings.py) |
 |  | [MistralAITextEmbedding](../../../semantic_kernel/connectors/ai/mistral_ai/services/mistral_ai_text_embedding.py) | ai_model_id, <br> api_key | MISTRALAI_EMBEDDING_MODEL_ID, <br> MISTRALAI_API_KEY | Yes, <br> Yes |  |
 | Ollama | [OllamaChatCompletion](../../../semantic_kernel/connectors/ai/ollama/services/ollama_chat_completion.py) | ai_model_id, <br> host | OLLAMA_CHAT_MODEL_ID, <br> OLLAMA_HOST | Yes, <br> No | [OllamaSettings](../../../semantic_kernel/connectors/ai/ollama/ollama_settings.py) |

python/samples/concepts/setup/chat_completion_services.py

@@ -28,6 +28,7 @@ class Services(str, Enum):
     ONNX = "onnx"
     VERTEX_AI = "vertex_ai"
     DEEPSEEK = "deepseek"
+    NVIDIA = "nvidia"
 
 
 service_id = "default"
@@ -64,6 +65,7 @@ def get_chat_completion_service_and_request_settings(
         Services.ONNX: lambda: get_onnx_chat_completion_service_and_request_settings(),
         Services.VERTEX_AI: lambda: get_vertex_ai_chat_completion_service_and_request_settings(),
         Services.DEEPSEEK: lambda: get_deepseek_chat_completion_service_and_request_settings(),
+        Services.NVIDIA: lambda: get_nvidia_chat_completion_service_and_request_settings(),
     }
 
     # Call the appropriate lambda or function based on the service name
@@ -414,3 +416,27 @@ def get_deepseek_chat_completion_service_and_request_settings() -> tuple[
     request_settings = OpenAIChatPromptExecutionSettings(service_id=service_id)
 
     return chat_service, request_settings
+
+
+def get_nvidia_chat_completion_service_and_request_settings() -> tuple[
+    "ChatCompletionClientBase", "PromptExecutionSettings"
+]:
+    """Return NVIDIA chat completion service and request settings.
+
+    The service credentials can be read by 3 ways:
+    1. Via the constructor
+    2. Via the environment variables
+    3. Via an environment file
+
+    The request settings control the behavior of the service. The default settings are sufficient to get started.
+    However, you can adjust the settings to suit your needs.
+    Note: Some of the settings are NOT meant to be set by the user.
+    Please refer to the Semantic Kernel Python documentation for more information:
+    https://learn.microsoft.com/en-us/python/api/semantic-kernel/semantic_kernel?view=semantic-kernel-python
+    """
+    from semantic_kernel.connectors.ai.nvidia import NvidiaChatCompletion, NvidiaChatPromptExecutionSettings
+
+    chat_service = NvidiaChatCompletion(service_id=service_id)
+    request_settings = NvidiaChatPromptExecutionSettings(service_id=service_id)
+
+    return chat_service, request_settings

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

@@ -1,6 +1,6 @@
 # semantic_kernel.connectors.ai.nvidia
 
-This connector enables integration with NVIDIA NIM API for text embeddings. It allows you to use NVIDIA's embedding models within the Semantic Kernel framework.
+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
 
@@ -13,6 +13,8 @@ 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
@@ -30,3 +32,35 @@ kernel.add_service(embedding_service)
 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

@@ -1,13 +1,17 @@
 # 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",

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

@@ -2,7 +2,7 @@
 
 from typing import Annotated, Any, Literal
 
-from pydantic import Field
+from pydantic import BaseModel, Field
 
 from semantic_kernel.connectors.ai.prompt_execution_settings import PromptExecutionSettings
 
@@ -13,18 +13,6 @@ class NvidiaPromptExecutionSettings(PromptExecutionSettings):
     format: Literal["json"] | None = None
     options: dict[str, Any] | None = None
 
-    def prepare_settings_dict(self, **kwargs) -> dict[str, Any]:
-        """Prepare the settings as a dictionary for sending to the AI service.
-
-        By default, this method excludes the service_id and extension_data fields.
-        As well as any fields that are None.
-        """
-        return self.model_dump(
-            exclude={"service_id", "extension_data", "structured_json_response", "input_type", "truncate"},
-            exclude_none=True,
-            by_alias=True,
-        )
-
 
 class NvidiaEmbeddingPromptExecutionSettings(NvidiaPromptExecutionSettings):
     """Settings for NVIDIA embedding prompt execution."""
@@ -39,3 +27,47 @@ class NvidiaEmbeddingPromptExecutionSettings(NvidiaPromptExecutionSettings):
     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,
+        )