mistralai 1.1.0__py3-none-any.whl → 1.2.0__py3-none-any.whl

mistralai_azure/chat.py

@@ -16,10 +16,9 @@ class Chat(BaseSDK):
         *,
         messages: Union[List[models.Messages], List[models.MessagesTypedDict]],
         model: OptionalNullable[str] = "azureai",
-        temperature: Optional[float] = 0.7,
+        temperature: OptionalNullable[float] = UNSET,
         top_p: Optional[float] = 1,
         max_tokens: OptionalNullable[int] = UNSET,
-        min_tokens: OptionalNullable[int] = UNSET,
         stream: Optional[bool] = True,
         stop: Optional[Union[models.Stop, models.StopTypedDict]] = None,
         random_seed: OptionalNullable[int] = UNSET,
@@ -35,6 +34,9 @@ class Chat(BaseSDK):
                 models.ChatCompletionStreamRequestToolChoiceTypedDict,
             ]
         ] = None,
+        presence_penalty: Optional[float] = 0,
+        frequency_penalty: Optional[float] = 0,
+        n: OptionalNullable[int] = UNSET,
         safe_prompt: Optional[bool] = False,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
@@ -46,16 +48,18 @@ class Chat(BaseSDK):
 
         :param messages: The prompt(s) to generate completions for, encoded as a list of dict with role and content.
         :param model: The ID of the model to use for this request.
-        :param temperature: What sampling temperature to use, between 0.0 and 1.0. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both.
+        :param temperature: What sampling temperature to use, we recommend between 0.0 and 0.7. Higher values like 0.7 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both. The default value varies depending on the model you are targeting. Call the `/models` endpoint to retrieve the appropriate value.
         :param top_p: Nucleus sampling, where the model considers the results of the tokens with `top_p` probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. We generally recommend altering this or `temperature` but not both.
         :param max_tokens: The maximum number of tokens to generate in the completion. The token count of your prompt plus `max_tokens` cannot exceed the model's context length.
-        :param min_tokens: The minimum number of tokens to generate in the completion.
         :param stream:
         :param stop: Stop generation if this token is detected. Or if one of these tokens is detected when providing an array
         :param random_seed: The seed to use for random sampling. If set, different calls will generate deterministic results.
         :param response_format:
         :param tools:
         :param tool_choice:
+        :param presence_penalty: presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
+        :param frequency_penalty: frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
+        :param n: Number of completions to return for each request, input tokens are only billed once.
         :param safe_prompt: Whether to inject a safety prompt before all conversations.
         :param retries: Override the default retry configuration for this method
         :param server_url: Override the default server URL for this method
@@ -74,7 +78,6 @@ class Chat(BaseSDK):
             temperature=temperature,
             top_p=top_p,
             max_tokens=max_tokens,
-            min_tokens=min_tokens,
             stream=stream,
             stop=stop,
             random_seed=random_seed,
@@ -86,6 +89,9 @@ class Chat(BaseSDK):
             tool_choice=utils.get_pydantic_model(
                 tool_choice, Optional[models.ChatCompletionStreamRequestToolChoice]
             ),
+            presence_penalty=presence_penalty,
+            frequency_penalty=frequency_penalty,
+            n=n,
             safe_prompt=safe_prompt,
         )
 
@@ -135,18 +141,21 @@ class Chat(BaseSDK):
                 sentinel="[DONE]",
             )
         if utils.match_response(http_res, "422", "application/json"):
-            data = utils.unmarshal_json(http_res.text, models.HTTPValidationErrorData)
+            http_res_text = utils.stream_to_text(http_res)
+            data = utils.unmarshal_json(http_res_text, models.HTTPValidationErrorData)
             raise models.HTTPValidationError(data=data)
         if utils.match_response(http_res, ["4XX", "5XX"], "*"):
+            http_res_text = utils.stream_to_text(http_res)
             raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res.text, http_res
+                "API error occurred", http_res.status_code, http_res_text, http_res
             )
 
         content_type = http_res.headers.get("Content-Type")
+        http_res_text = utils.stream_to_text(http_res)
         raise models.SDKError(
             f"Unexpected response received (code: {http_res.status_code}, type: {content_type})",
             http_res.status_code,
-            http_res.text,
+            http_res_text,
             http_res,
         )
 
@@ -155,10 +164,9 @@ class Chat(BaseSDK):
         *,
         messages: Union[List[models.Messages], List[models.MessagesTypedDict]],
         model: OptionalNullable[str] = "azureai",
-        temperature: Optional[float] = 0.7,
+        temperature: OptionalNullable[float] = UNSET,
         top_p: Optional[float] = 1,
         max_tokens: OptionalNullable[int] = UNSET,
-        min_tokens: OptionalNullable[int] = UNSET,
         stream: Optional[bool] = True,
         stop: Optional[Union[models.Stop, models.StopTypedDict]] = None,
         random_seed: OptionalNullable[int] = UNSET,
@@ -174,6 +182,9 @@ class Chat(BaseSDK):
                 models.ChatCompletionStreamRequestToolChoiceTypedDict,
             ]
         ] = None,
+        presence_penalty: Optional[float] = 0,
+        frequency_penalty: Optional[float] = 0,
+        n: OptionalNullable[int] = UNSET,
         safe_prompt: Optional[bool] = False,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
@@ -185,16 +196,18 @@ class Chat(BaseSDK):
 
         :param messages: The prompt(s) to generate completions for, encoded as a list of dict with role and content.
         :param model: The ID of the model to use for this request.
-        :param temperature: What sampling temperature to use, between 0.0 and 1.0. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both.
+        :param temperature: What sampling temperature to use, we recommend between 0.0 and 0.7. Higher values like 0.7 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both. The default value varies depending on the model you are targeting. Call the `/models` endpoint to retrieve the appropriate value.
         :param top_p: Nucleus sampling, where the model considers the results of the tokens with `top_p` probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. We generally recommend altering this or `temperature` but not both.
         :param max_tokens: The maximum number of tokens to generate in the completion. The token count of your prompt plus `max_tokens` cannot exceed the model's context length.
-        :param min_tokens: The minimum number of tokens to generate in the completion.
         :param stream:
         :param stop: Stop generation if this token is detected. Or if one of these tokens is detected when providing an array
         :param random_seed: The seed to use for random sampling. If set, different calls will generate deterministic results.
         :param response_format:
         :param tools:
         :param tool_choice:
+        :param presence_penalty: presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
+        :param frequency_penalty: frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
+        :param n: Number of completions to return for each request, input tokens are only billed once.
         :param safe_prompt: Whether to inject a safety prompt before all conversations.
         :param retries: Override the default retry configuration for this method
         :param server_url: Override the default server URL for this method
@@ -213,7 +226,6 @@ class Chat(BaseSDK):
             temperature=temperature,
             top_p=top_p,
             max_tokens=max_tokens,
-            min_tokens=min_tokens,
             stream=stream,
             stop=stop,
             random_seed=random_seed,
@@ -225,6 +237,9 @@ class Chat(BaseSDK):
             tool_choice=utils.get_pydantic_model(
                 tool_choice, Optional[models.ChatCompletionStreamRequestToolChoice]
             ),
+            presence_penalty=presence_penalty,
+            frequency_penalty=frequency_penalty,
+            n=n,
             safe_prompt=safe_prompt,
         )
 
@@ -274,18 +289,21 @@ class Chat(BaseSDK):
                 sentinel="[DONE]",
             )
         if utils.match_response(http_res, "422", "application/json"):
-            data = utils.unmarshal_json(http_res.text, models.HTTPValidationErrorData)
+            http_res_text = await utils.stream_to_text_async(http_res)
+            data = utils.unmarshal_json(http_res_text, models.HTTPValidationErrorData)
             raise models.HTTPValidationError(data=data)
         if utils.match_response(http_res, ["4XX", "5XX"], "*"):
+            http_res_text = await utils.stream_to_text_async(http_res)
             raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res.text, http_res
+                "API error occurred", http_res.status_code, http_res_text, http_res
             )
 
         content_type = http_res.headers.get("Content-Type")
+        http_res_text = await utils.stream_to_text_async(http_res)
         raise models.SDKError(
             f"Unexpected response received (code: {http_res.status_code}, type: {content_type})",
             http_res.status_code,
-            http_res.text,
+            http_res_text,
             http_res,
         )
 
@@ -297,10 +315,9 @@ class Chat(BaseSDK):
             List[models.ChatCompletionRequestMessagesTypedDict],
         ],
         model: OptionalNullable[str] = "azureai",
-        temperature: Optional[float] = 0.7,
+        temperature: OptionalNullable[float] = UNSET,
         top_p: Optional[float] = 1,
         max_tokens: OptionalNullable[int] = UNSET,
-        min_tokens: OptionalNullable[int] = UNSET,
         stream: Optional[bool] = False,
         stop: Optional[
             Union[
@@ -321,6 +338,9 @@ class Chat(BaseSDK):
                 models.ChatCompletionRequestToolChoiceTypedDict,
             ]
         ] = None,
+        presence_penalty: Optional[float] = 0,
+        frequency_penalty: Optional[float] = 0,
+        n: OptionalNullable[int] = UNSET,
         safe_prompt: Optional[bool] = False,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
@@ -330,16 +350,18 @@ class Chat(BaseSDK):
 
         :param messages: The prompt(s) to generate completions for, encoded as a list of dict with role and content.
         :param model: The ID of the model to use for this request.
-        :param temperature: What sampling temperature to use, between 0.0 and 1.0. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both.
+        :param temperature: What sampling temperature to use, we recommend between 0.0 and 0.7. Higher values like 0.7 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both. The default value varies depending on the model you are targeting. Call the `/models` endpoint to retrieve the appropriate value.
         :param top_p: Nucleus sampling, where the model considers the results of the tokens with `top_p` probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. We generally recommend altering this or `temperature` but not both.
         :param max_tokens: The maximum number of tokens to generate in the completion. The token count of your prompt plus `max_tokens` cannot exceed the model's context length.
-        :param min_tokens: The minimum number of tokens to generate in the completion.
         :param stream: Whether to stream back partial progress. If set, tokens will be sent as data-only server-side events as they become available, with the stream terminated by a data: [DONE] message. Otherwise, the server will hold the request open until the timeout or until completion, with the response containing the full result as JSON.
         :param stop: Stop generation if this token is detected. Or if one of these tokens is detected when providing an array
         :param random_seed: The seed to use for random sampling. If set, different calls will generate deterministic results.
         :param response_format:
         :param tools:
         :param tool_choice:
+        :param presence_penalty: presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
+        :param frequency_penalty: frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
+        :param n: Number of completions to return for each request, input tokens are only billed once.
         :param safe_prompt: Whether to inject a safety prompt before all conversations.
         :param retries: Override the default retry configuration for this method
         :param server_url: Override the default server URL for this method
@@ -358,7 +380,6 @@ class Chat(BaseSDK):
             temperature=temperature,
             top_p=top_p,
             max_tokens=max_tokens,
-            min_tokens=min_tokens,
             stream=stream,
             stop=stop,
             random_seed=random_seed,
@@ -372,6 +393,9 @@ class Chat(BaseSDK):
             tool_choice=utils.get_pydantic_model(
                 tool_choice, Optional[models.ChatCompletionRequestToolChoice]
             ),
+            presence_penalty=presence_penalty,
+            frequency_penalty=frequency_penalty,
+            n=n,
             safe_prompt=safe_prompt,
         )
 
@@ -421,15 +445,17 @@ class Chat(BaseSDK):
             data = utils.unmarshal_json(http_res.text, models.HTTPValidationErrorData)
             raise models.HTTPValidationError(data=data)
         if utils.match_response(http_res, ["4XX", "5XX"], "*"):
+            http_res_text = utils.stream_to_text(http_res)
             raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res.text, http_res
+                "API error occurred", http_res.status_code, http_res_text, http_res
             )
 
         content_type = http_res.headers.get("Content-Type")
+        http_res_text = utils.stream_to_text(http_res)
         raise models.SDKError(
             f"Unexpected response received (code: {http_res.status_code}, type: {content_type})",
             http_res.status_code,
-            http_res.text,
+            http_res_text,
             http_res,
         )
 
@@ -441,10 +467,9 @@ class Chat(BaseSDK):
             List[models.ChatCompletionRequestMessagesTypedDict],
         ],
         model: OptionalNullable[str] = "azureai",
-        temperature: Optional[float] = 0.7,
+        temperature: OptionalNullable[float] = UNSET,
         top_p: Optional[float] = 1,
         max_tokens: OptionalNullable[int] = UNSET,
-        min_tokens: OptionalNullable[int] = UNSET,
         stream: Optional[bool] = False,
         stop: Optional[
             Union[
@@ -465,6 +490,9 @@ class Chat(BaseSDK):
                 models.ChatCompletionRequestToolChoiceTypedDict,
             ]
         ] = None,
+        presence_penalty: Optional[float] = 0,
+        frequency_penalty: Optional[float] = 0,
+        n: OptionalNullable[int] = UNSET,
         safe_prompt: Optional[bool] = False,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
@@ -474,16 +502,18 @@ class Chat(BaseSDK):
 
         :param messages: The prompt(s) to generate completions for, encoded as a list of dict with role and content.
         :param model: The ID of the model to use for this request.
-        :param temperature: What sampling temperature to use, between 0.0 and 1.0. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both.
+        :param temperature: What sampling temperature to use, we recommend between 0.0 and 0.7. Higher values like 0.7 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both. The default value varies depending on the model you are targeting. Call the `/models` endpoint to retrieve the appropriate value.
         :param top_p: Nucleus sampling, where the model considers the results of the tokens with `top_p` probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. We generally recommend altering this or `temperature` but not both.
         :param max_tokens: The maximum number of tokens to generate in the completion. The token count of your prompt plus `max_tokens` cannot exceed the model's context length.
-        :param min_tokens: The minimum number of tokens to generate in the completion.
         :param stream: Whether to stream back partial progress. If set, tokens will be sent as data-only server-side events as they become available, with the stream terminated by a data: [DONE] message. Otherwise, the server will hold the request open until the timeout or until completion, with the response containing the full result as JSON.
         :param stop: Stop generation if this token is detected. Or if one of these tokens is detected when providing an array
         :param random_seed: The seed to use for random sampling. If set, different calls will generate deterministic results.
         :param response_format:
         :param tools:
         :param tool_choice:
+        :param presence_penalty: presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
+        :param frequency_penalty: frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
+        :param n: Number of completions to return for each request, input tokens are only billed once.
         :param safe_prompt: Whether to inject a safety prompt before all conversations.
         :param retries: Override the default retry configuration for this method
         :param server_url: Override the default server URL for this method
@@ -502,7 +532,6 @@ class Chat(BaseSDK):
             temperature=temperature,
             top_p=top_p,
             max_tokens=max_tokens,
-            min_tokens=min_tokens,
             stream=stream,
             stop=stop,
             random_seed=random_seed,
@@ -516,6 +545,9 @@ class Chat(BaseSDK):
             tool_choice=utils.get_pydantic_model(
                 tool_choice, Optional[models.ChatCompletionRequestToolChoice]
             ),
+            presence_penalty=presence_penalty,
+            frequency_penalty=frequency_penalty,
+            n=n,
             safe_prompt=safe_prompt,
         )
 
@@ -565,14 +597,16 @@ class Chat(BaseSDK):
             data = utils.unmarshal_json(http_res.text, models.HTTPValidationErrorData)
             raise models.HTTPValidationError(data=data)
         if utils.match_response(http_res, ["4XX", "5XX"], "*"):
+            http_res_text = await utils.stream_to_text_async(http_res)
             raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res.text, http_res
+                "API error occurred", http_res.status_code, http_res_text, http_res
             )
 
         content_type = http_res.headers.get("Content-Type")
+        http_res_text = await utils.stream_to_text_async(http_res)
         raise models.SDKError(
             f"Unexpected response received (code: {http_res.status_code}, type: {content_type})",
             http_res.status_code,
-            http_res.text,
+            http_res_text,
             http_res,
         )

mistralai_azure/models/__init__.py

@@ -2,6 +2,8 @@
 
 from .assistantmessage import (
     AssistantMessage,
+    AssistantMessageContent,
+    AssistantMessageContentTypedDict,
     AssistantMessageRole,
     AssistantMessageTypedDict,
 )
@@ -42,7 +44,7 @@ from .completionresponsestreamchoice import (
     FinishReason,
 )
 from .contentchunk import ContentChunk, ContentChunkTypedDict
-from .deltamessage import DeltaMessage, DeltaMessageTypedDict
+from .deltamessage import Content, ContentTypedDict, DeltaMessage, DeltaMessageTypedDict
 from .function import Function, FunctionTypedDict
 from .functioncall import (
     Arguments,
@@ -57,10 +59,10 @@ from .responseformats import ResponseFormats
 from .sdkerror import SDKError
 from .security import Security, SecurityTypedDict
 from .systemmessage import (
-    Content,
-    ContentTypedDict,
     Role,
     SystemMessage,
+    SystemMessageContent,
+    SystemMessageContentTypedDict,
     SystemMessageTypedDict,
 )
 from .textchunk import TextChunk, TextChunkTypedDict, Type
@@ -89,6 +91,8 @@ __all__ = [
     "Arguments",
     "ArgumentsTypedDict",
     "AssistantMessage",
+    "AssistantMessageContent",
+    "AssistantMessageContentTypedDict",
     "AssistantMessageRole",
     "AssistantMessageTypedDict",
     "ChatCompletionChoice",
@@ -143,6 +147,8 @@ __all__ = [
     "Stop",
     "StopTypedDict",
     "SystemMessage",
+    "SystemMessageContent",
+    "SystemMessageContentTypedDict",
     "SystemMessageTypedDict",
     "TextChunk",
     "TextChunkTypedDict",

mistralai_azure/models/assistantmessage.py

@@ -1,6 +1,7 @@
 """Code generated by Speakeasy (https://speakeasy.com). DO NOT EDIT."""
 
 from __future__ import annotations
+from .contentchunk import ContentChunk, ContentChunkTypedDict
 from .toolcall import ToolCall, ToolCallTypedDict
 from mistralai_azure.types import (
     BaseModel,
@@ -10,28 +11,32 @@ from mistralai_azure.types import (
     UNSET_SENTINEL,
 )
 from pydantic import model_serializer
-from typing import List, Literal, Optional, TypedDict
-from typing_extensions import NotRequired
+from typing import List, Literal, Optional, Union
+from typing_extensions import NotRequired, TypedDict
+
+
+AssistantMessageContentTypedDict = Union[str, List[ContentChunkTypedDict]]
+
+
+AssistantMessageContent = Union[str, List[ContentChunk]]
 
 
 AssistantMessageRole = Literal["assistant"]
 
 
 class AssistantMessageTypedDict(TypedDict):
-    content: NotRequired[Nullable[str]]
+    content: NotRequired[Nullable[AssistantMessageContentTypedDict]]
     tool_calls: NotRequired[Nullable[List[ToolCallTypedDict]]]
     prefix: NotRequired[bool]
-    r"""Set this to `true` when adding an assistant message as prefix to condition the model response. The role of the prefix message is to force the model to start its answer by the content of the message."""
     role: NotRequired[AssistantMessageRole]
 
 
 class AssistantMessage(BaseModel):
-    content: OptionalNullable[str] = UNSET
+    content: OptionalNullable[AssistantMessageContent] = UNSET
 
     tool_calls: OptionalNullable[List[ToolCall]] = UNSET
 
     prefix: Optional[bool] = False
-    r"""Set this to `true` when adding an assistant message as prefix to condition the model response. The role of the prefix message is to force the model to start its answer by the content of the message."""
 
     role: Optional[AssistantMessageRole] = "assistant"
 

mistralai_azure/models/chatcompletionchoice.py

@@ -2,12 +2,15 @@
 
 from __future__ import annotations
 from .assistantmessage import AssistantMessage, AssistantMessageTypedDict
-from mistralai_azure.types import BaseModel
-from typing import Literal, TypedDict
+from mistralai_azure.types import BaseModel, UnrecognizedStr
+from mistralai_azure.utils import validate_open_enum
+from pydantic.functional_validators import PlainValidator
+from typing import Literal, Union
+from typing_extensions import Annotated, TypedDict
 
 
-ChatCompletionChoiceFinishReason = Literal[
-    "stop", "length", "model_length", "error", "tool_calls"
+ChatCompletionChoiceFinishReason = Union[
+    Literal["stop", "length", "model_length", "error", "tool_calls"], UnrecognizedStr
 ]
 
 
@@ -22,4 +25,6 @@ class ChatCompletionChoice(BaseModel):
 
     message: AssistantMessage
 
-    finish_reason: ChatCompletionChoiceFinishReason
+    finish_reason: Annotated[
+        ChatCompletionChoiceFinishReason, PlainValidator(validate_open_enum(False))
+    ]

mistralai_azure/models/chatcompletionrequest.py

@@ -18,8 +18,8 @@ from mistralai_azure.types import (
 )
 from mistralai_azure.utils import get_discriminator
 from pydantic import Discriminator, Tag, model_serializer
-from typing import List, Optional, TypedDict, Union
-from typing_extensions import Annotated, NotRequired
+from typing import List, Optional, Union
+from typing_extensions import Annotated, NotRequired, TypedDict
 
 
 ChatCompletionRequestStopTypedDict = Union[str, List[str]]
@@ -60,14 +60,12 @@ class ChatCompletionRequestTypedDict(TypedDict):
     r"""The prompt(s) to generate completions for, encoded as a list of dict with role and content."""
     model: NotRequired[Nullable[str]]
     r"""The ID of the model to use for this request."""
-    temperature: NotRequired[float]
-    r"""What sampling temperature to use, between 0.0 and 1.0. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both."""
+    temperature: NotRequired[Nullable[float]]
+    r"""What sampling temperature to use, we recommend between 0.0 and 0.7. Higher values like 0.7 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both. The default value varies depending on the model you are targeting. Call the `/models` endpoint to retrieve the appropriate value."""
     top_p: NotRequired[float]
     r"""Nucleus sampling, where the model considers the results of the tokens with `top_p` probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. We generally recommend altering this or `temperature` but not both."""
     max_tokens: NotRequired[Nullable[int]]
     r"""The maximum number of tokens to generate in the completion. The token count of your prompt plus `max_tokens` cannot exceed the model's context length."""
-    min_tokens: NotRequired[Nullable[int]]
-    r"""The minimum number of tokens to generate in the completion."""
     stream: NotRequired[bool]
     r"""Whether to stream back partial progress. If set, tokens will be sent as data-only server-side events as they become available, with the stream terminated by a data: [DONE] message. Otherwise, the server will hold the request open until the timeout or until completion, with the response containing the full result as JSON."""
     stop: NotRequired[ChatCompletionRequestStopTypedDict]
@@ -77,6 +75,12 @@ class ChatCompletionRequestTypedDict(TypedDict):
     response_format: NotRequired[ResponseFormatTypedDict]
     tools: NotRequired[Nullable[List[ToolTypedDict]]]
     tool_choice: NotRequired[ChatCompletionRequestToolChoiceTypedDict]
+    presence_penalty: NotRequired[float]
+    r"""presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative."""
+    frequency_penalty: NotRequired[float]
+    r"""frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition."""
+    n: NotRequired[Nullable[int]]
+    r"""Number of completions to return for each request, input tokens are only billed once."""
     safe_prompt: NotRequired[bool]
     r"""Whether to inject a safety prompt before all conversations."""
 
@@ -88,8 +92,8 @@ class ChatCompletionRequest(BaseModel):
     model: OptionalNullable[str] = "azureai"
     r"""The ID of the model to use for this request."""
 
-    temperature: Optional[float] = 0.7
-    r"""What sampling temperature to use, between 0.0 and 1.0. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both."""
+    temperature: OptionalNullable[float] = UNSET
+    r"""What sampling temperature to use, we recommend between 0.0 and 0.7. Higher values like 0.7 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both. The default value varies depending on the model you are targeting. Call the `/models` endpoint to retrieve the appropriate value."""
 
     top_p: Optional[float] = 1
     r"""Nucleus sampling, where the model considers the results of the tokens with `top_p` probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. We generally recommend altering this or `temperature` but not both."""
@@ -97,9 +101,6 @@ class ChatCompletionRequest(BaseModel):
     max_tokens: OptionalNullable[int] = UNSET
     r"""The maximum number of tokens to generate in the completion. The token count of your prompt plus `max_tokens` cannot exceed the model's context length."""
 
-    min_tokens: OptionalNullable[int] = UNSET
-    r"""The minimum number of tokens to generate in the completion."""
-
     stream: Optional[bool] = False
     r"""Whether to stream back partial progress. If set, tokens will be sent as data-only server-side events as they become available, with the stream terminated by a data: [DONE] message. Otherwise, the server will hold the request open until the timeout or until completion, with the response containing the full result as JSON."""
 
@@ -115,6 +116,15 @@ class ChatCompletionRequest(BaseModel):
 
     tool_choice: Optional[ChatCompletionRequestToolChoice] = None
 
+    presence_penalty: Optional[float] = 0
+    r"""presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative."""
+
+    frequency_penalty: Optional[float] = 0
+    r"""frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition."""
+
+    n: OptionalNullable[int] = UNSET
+    r"""Number of completions to return for each request, input tokens are only billed once."""
+
     safe_prompt: Optional[bool] = False
     r"""Whether to inject a safety prompt before all conversations."""
 
@@ -125,16 +135,25 @@ class ChatCompletionRequest(BaseModel):
             "temperature",
             "top_p",
             "max_tokens",
-            "min_tokens",
             "stream",
             "stop",
             "random_seed",
             "response_format",
             "tools",
             "tool_choice",
+            "presence_penalty",
+            "frequency_penalty",
+            "n",
             "safe_prompt",
         ]
-        nullable_fields = ["model", "max_tokens", "min_tokens", "random_seed", "tools"]
+        nullable_fields = [
+            "model",
+            "temperature",
+            "max_tokens",
+            "random_seed",
+            "tools",
+            "n",
+        ]
         null_default_fields = []
 
         serialized = handler(self)

mistralai_azure/models/chatcompletionresponse.py

@@ -4,8 +4,8 @@ from __future__ import annotations
 from .chatcompletionchoice import ChatCompletionChoice, ChatCompletionChoiceTypedDict
 from .usageinfo import UsageInfo, UsageInfoTypedDict
 from mistralai_azure.types import BaseModel
-from typing import List, Optional, TypedDict
-from typing_extensions import NotRequired
+from typing import List, Optional
+from typing_extensions import NotRequired, TypedDict
 
 
 class ChatCompletionResponseTypedDict(TypedDict):