mistralai 1.9.11__py3-none-any.whl → 1.10.0__py3-none-any.whl

mistralai/_hooks/registration.py

@@ -1,5 +1,6 @@
 from .custom_user_agent import CustomUserAgentHook
 from .deprecation_warning import DeprecationWarningHook
+from .tracing import TracingHook
 from .types import Hooks
 
 # This file is only ever generated once on the first generation and then is free to be modified.
@@ -13,5 +14,9 @@ def init_hooks(hooks: Hooks):
     with an instance of a hook that implements that specific Hook interface
     Hooks are registered per SDK instance, and are valid for the lifetime of the SDK instance
     """
+    tracing_hook = TracingHook()
     hooks.register_before_request_hook(CustomUserAgentHook())
     hooks.register_after_success_hook(DeprecationWarningHook())
+    hooks.register_after_success_hook(tracing_hook)
+    hooks.register_before_request_hook(tracing_hook)
+    hooks.register_after_error_hook(tracing_hook)

mistralai/_hooks/tracing.py

@@ -0,0 +1,50 @@
+import logging
+from typing import Optional, Tuple, Union
+
+import httpx
+from opentelemetry.trace import Span
+
+from ..extra.observability.otel import (
+    get_or_create_otel_tracer,
+    get_response_and_error,
+    get_traced_request_and_span,
+    get_traced_response,
+)
+from .types import (
+    AfterErrorContext,
+    AfterErrorHook,
+    AfterSuccessContext,
+    AfterSuccessHook,
+    BeforeRequestContext,
+    BeforeRequestHook,
+)
+
+logger = logging.getLogger(__name__)
+
+
+class TracingHook(BeforeRequestHook, AfterSuccessHook, AfterErrorHook):
+    def __init__(self) -> None:
+        self.tracing_enabled, self.tracer = get_or_create_otel_tracer()
+        self.request_span: Optional[Span] = None
+
+    def before_request(
+        self, hook_ctx: BeforeRequestContext, request: httpx.Request
+    ) -> Union[httpx.Request, Exception]:
+        request, self.request_span = get_traced_request_and_span(tracing_enabled=self.tracing_enabled, tracer=self.tracer, span=self.request_span, operation_id=hook_ctx.operation_id, request=request)
+        return request
+
+    def after_success(
+        self, hook_ctx: AfterSuccessContext, response: httpx.Response
+    ) -> Union[httpx.Response, Exception]:
+        response = get_traced_response(tracing_enabled=self.tracing_enabled, tracer=self.tracer, span=self.request_span, operation_id=hook_ctx.operation_id, response=response)
+        return response
+
+    def after_error(
+        self,
+        hook_ctx: AfterErrorContext,
+        response: Optional[httpx.Response],
+        error: Optional[Exception],
+    ) -> Union[Tuple[Optional[httpx.Response], Optional[Exception]], Exception]:
+        if response:
+            response, error = get_response_and_error(tracing_enabled=self.tracing_enabled, tracer=self.tracer, span=self.request_span, operation_id=hook_ctx.operation_id, response=response, error=error)
+        return response, error

mistralai/_version.py

@@ -3,10 +3,10 @@
 import importlib.metadata
 
 __title__: str = "mistralai"
-__version__: str = "1.9.11"
+__version__: str = "1.10.0"
 __openapi_doc_version__: str = "1.0.0"
 __gen_version__: str = "2.687.13"
-__user_agent__: str = "speakeasy-sdk/python 1.9.11 2.687.13 1.0.0 mistralai"
+__user_agent__: str = "speakeasy-sdk/python 1.10.0 2.687.13 1.0.0 mistralai"
 
 try:
     if __package__ is not None:

mistralai/accesses.py

@@ -194,10 +194,10 @@ class Accesses(BaseSDK):
         self,
         *,
         library_id: str,
-        org_id: str,
         level: models.ShareEnum,
         share_with_uuid: str,
         share_with_type: models.EntityType,
+        org_id: OptionalNullable[str] = UNSET,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -208,10 +208,10 @@ class Accesses(BaseSDK):
         Given a library id, you can create or update the access level of an entity. You have to be owner of the library to share a library. An owner cannot change their own role. A library cannot be shared outside of the organization.
 
         :param library_id:
-        :param org_id:
         :param level:
         :param share_with_uuid: The id of the entity (user, workspace or organization) to share with
         :param share_with_type: The type of entity, used to share a library.
+        :param org_id:
         :param retries: Override the default retry configuration for this method
         :param server_url: Override the default server URL for this method
         :param timeout_ms: Override the default request timeout configuration for this method in milliseconds
@@ -300,10 +300,10 @@ class Accesses(BaseSDK):
         self,
         *,
         library_id: str,
-        org_id: str,
         level: models.ShareEnum,
         share_with_uuid: str,
         share_with_type: models.EntityType,
+        org_id: OptionalNullable[str] = UNSET,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -314,10 +314,10 @@ class Accesses(BaseSDK):
         Given a library id, you can create or update the access level of an entity. You have to be owner of the library to share a library. An owner cannot change their own role. A library cannot be shared outside of the organization.
 
         :param library_id:
-        :param org_id:
         :param level:
         :param share_with_uuid: The id of the entity (user, workspace or organization) to share with
         :param share_with_type: The type of entity, used to share a library.
+        :param org_id:
         :param retries: Override the default retry configuration for this method
         :param server_url: Override the default server URL for this method
         :param timeout_ms: Override the default request timeout configuration for this method in milliseconds
@@ -406,9 +406,9 @@ class Accesses(BaseSDK):
         self,
         *,
         library_id: str,
-        org_id: str,
         share_with_uuid: str,
         share_with_type: models.EntityType,
+        org_id: OptionalNullable[str] = UNSET,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -419,9 +419,9 @@ class Accesses(BaseSDK):
         Given a library id, you can delete the access level of an entity. An owner cannot delete it's own access. You have to be the owner of the library to delete an acces other than yours.
 
         :param library_id:
-        :param org_id:
         :param share_with_uuid: The id of the entity (user, workspace or organization) to share with
         :param share_with_type: The type of entity, used to share a library.
+        :param org_id:
         :param retries: Override the default retry configuration for this method
         :param server_url: Override the default server URL for this method
         :param timeout_ms: Override the default request timeout configuration for this method in milliseconds
@@ -509,9 +509,9 @@ class Accesses(BaseSDK):
         self,
         *,
         library_id: str,
-        org_id: str,
         share_with_uuid: str,
         share_with_type: models.EntityType,
+        org_id: OptionalNullable[str] = UNSET,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -522,9 +522,9 @@ class Accesses(BaseSDK):
         Given a library id, you can delete the access level of an entity. An owner cannot delete it's own access. You have to be the owner of the library to delete an acces other than yours.
 
         :param library_id:
-        :param org_id:
         :param share_with_uuid: The id of the entity (user, workspace or organization) to share with
         :param share_with_type: The type of entity, used to share a library.
+        :param org_id:
         :param retries: Override the default retry configuration for this method
         :param server_url: Override the default server URL for this method
         :param timeout_ms: Override the default request timeout configuration for this method in milliseconds

mistralai/agents.py

@@ -6,7 +6,7 @@ from mistralai._hooks import HookContext
 from mistralai.types import OptionalNullable, UNSET
 from mistralai.utils import eventstreaming, get_security_from_env
 from mistralai.utils.unmarshal_json_response import unmarshal_json_response
-from typing import Any, List, Mapping, Optional, Union
+from typing import Any, Dict, List, Mapping, Optional, Union
 
 
 class Agents(BaseSDK):
@@ -29,6 +29,7 @@ class Agents(BaseSDK):
             ]
         ] = None,
         random_seed: OptionalNullable[int] = UNSET,
+        metadata: OptionalNullable[Dict[str, Any]] = UNSET,
         response_format: Optional[
             Union[models.ResponseFormat, models.ResponseFormatTypedDict]
         ] = None,
@@ -62,13 +63,14 @@ class Agents(BaseSDK):
         :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 metadata:
+        :param response_format: Specify the format that the model must output. By default it will use `{ \"type\": \"text\" }`. Setting to `{ \"type\": \"json_object\" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ \"type\": \"json_schema\" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide.
         :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 presence_penalty: The `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: The `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 prediction:
+        :param prediction: Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content.
         :param parallel_tool_calls:
         :param prompt_mode: Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used.
         :param retries: Override the default retry configuration for this method
@@ -91,6 +93,7 @@ class Agents(BaseSDK):
             stream=stream,
             stop=stop,
             random_seed=random_seed,
+            metadata=metadata,
             messages=utils.get_pydantic_model(
                 messages, List[models.AgentsCompletionRequestMessages]
             ),
@@ -188,6 +191,7 @@ class Agents(BaseSDK):
             ]
         ] = None,
         random_seed: OptionalNullable[int] = UNSET,
+        metadata: OptionalNullable[Dict[str, Any]] = UNSET,
         response_format: Optional[
             Union[models.ResponseFormat, models.ResponseFormatTypedDict]
         ] = None,
@@ -221,13 +225,14 @@ class Agents(BaseSDK):
         :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 metadata:
+        :param response_format: Specify the format that the model must output. By default it will use `{ \"type\": \"text\" }`. Setting to `{ \"type\": \"json_object\" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ \"type\": \"json_schema\" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide.
         :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 presence_penalty: The `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: The `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 prediction:
+        :param prediction: Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content.
         :param parallel_tool_calls:
         :param prompt_mode: Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used.
         :param retries: Override the default retry configuration for this method
@@ -250,6 +255,7 @@ class Agents(BaseSDK):
             stream=stream,
             stop=stop,
             random_seed=random_seed,
+            metadata=metadata,
             messages=utils.get_pydantic_model(
                 messages, List[models.AgentsCompletionRequestMessages]
             ),
@@ -347,6 +353,7 @@ class Agents(BaseSDK):
             ]
         ] = None,
         random_seed: OptionalNullable[int] = UNSET,
+        metadata: OptionalNullable[Dict[str, Any]] = UNSET,
         response_format: Optional[
             Union[models.ResponseFormat, models.ResponseFormatTypedDict]
         ] = None,
@@ -382,13 +389,14 @@ class Agents(BaseSDK):
         :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 metadata:
+        :param response_format: Specify the format that the model must output. By default it will use `{ \"type\": \"text\" }`. Setting to `{ \"type\": \"json_object\" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ \"type\": \"json_schema\" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide.
         :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 presence_penalty: The `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: The `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 prediction:
+        :param prediction: Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content.
         :param parallel_tool_calls:
         :param prompt_mode: Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used.
         :param retries: Override the default retry configuration for this method
@@ -411,6 +419,7 @@ class Agents(BaseSDK):
             stream=stream,
             stop=stop,
             random_seed=random_seed,
+            metadata=metadata,
             messages=utils.get_pydantic_model(
                 messages, List[models.AgentsCompletionStreamRequestMessages]
             ),
@@ -516,6 +525,7 @@ class Agents(BaseSDK):
             ]
         ] = None,
         random_seed: OptionalNullable[int] = UNSET,
+        metadata: OptionalNullable[Dict[str, Any]] = UNSET,
         response_format: Optional[
             Union[models.ResponseFormat, models.ResponseFormatTypedDict]
         ] = None,
@@ -551,13 +561,14 @@ class Agents(BaseSDK):
         :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 metadata:
+        :param response_format: Specify the format that the model must output. By default it will use `{ \"type\": \"text\" }`. Setting to `{ \"type\": \"json_object\" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ \"type\": \"json_schema\" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide.
         :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 presence_penalty: The `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: The `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 prediction:
+        :param prediction: Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content.
         :param parallel_tool_calls:
         :param prompt_mode: Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used.
         :param retries: Override the default retry configuration for this method
@@ -580,6 +591,7 @@ class Agents(BaseSDK):
             stream=stream,
             stop=stop,
             random_seed=random_seed,
+            metadata=metadata,
             messages=utils.get_pydantic_model(
                 messages, List[models.AgentsCompletionStreamRequestMessages]
             ),

mistralai/chat.py

@@ -6,7 +6,7 @@ from mistralai._hooks import HookContext
 from mistralai.types import OptionalNullable, UNSET
 from mistralai.utils import eventstreaming, get_security_from_env
 from mistralai.utils.unmarshal_json_response import unmarshal_json_response
-from typing import Any, List, Mapping, Optional, Union
+from typing import Any, Dict, List, Mapping, Optional, Union
 
 # region imports
 from typing import Type
@@ -105,6 +105,7 @@ class Chat(BaseSDK):
         stream: Optional[bool] = False,
         stop: Optional[Union[models.Stop, models.StopTypedDict]] = None,
         random_seed: OptionalNullable[int] = UNSET,
+        metadata: OptionalNullable[Dict[str, Any]] = UNSET,
         response_format: Optional[
             Union[models.ResponseFormat, models.ResponseFormatTypedDict]
         ] = None,
@@ -141,14 +142,15 @@ class Chat(BaseSDK):
         :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 metadata:
+        :param response_format: Specify the format that the model must output. By default it will use `{ \"type\": \"text\" }`. Setting to `{ \"type\": \"json_object\" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ \"type\": \"json_schema\" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide.
+        :param tools: A list of tools the model may call. Use this to provide a list of functions the model may generate JSON inputs for.
+        :param tool_choice: Controls which (if any) tool is called by the model. `none` means the model will not call any tool and instead generates a message. `auto` means the model can pick between generating a message or calling one or more tools. `any` or `required` means the model must call one or more tools. Specifying a particular tool via `{\"type\": \"function\", \"function\": {\"name\": \"my_function\"}}` forces the model to call that tool.
+        :param presence_penalty: The `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: The `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 prediction:
-        :param parallel_tool_calls:
+        :param prediction: Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content.
+        :param parallel_tool_calls: Whether to enable parallel function calling during tool use, when enabled the model can call multiple tools in parallel.
         :param prompt_mode: Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used.
         :param safe_prompt: Whether to inject a safety prompt before all conversations.
         :param retries: Override the default retry configuration for this method
@@ -174,6 +176,7 @@ class Chat(BaseSDK):
             stream=stream,
             stop=stop,
             random_seed=random_seed,
+            metadata=metadata,
             messages=utils.get_pydantic_model(messages, List[models.Messages]),
             response_format=utils.get_pydantic_model(
                 response_format, Optional[models.ResponseFormat]
@@ -263,6 +266,7 @@ class Chat(BaseSDK):
         stream: Optional[bool] = False,
         stop: Optional[Union[models.Stop, models.StopTypedDict]] = None,
         random_seed: OptionalNullable[int] = UNSET,
+        metadata: OptionalNullable[Dict[str, Any]] = UNSET,
         response_format: Optional[
             Union[models.ResponseFormat, models.ResponseFormatTypedDict]
         ] = None,
@@ -299,14 +303,15 @@ class Chat(BaseSDK):
         :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 metadata:
+        :param response_format: Specify the format that the model must output. By default it will use `{ \"type\": \"text\" }`. Setting to `{ \"type\": \"json_object\" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ \"type\": \"json_schema\" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide.
+        :param tools: A list of tools the model may call. Use this to provide a list of functions the model may generate JSON inputs for.
+        :param tool_choice: Controls which (if any) tool is called by the model. `none` means the model will not call any tool and instead generates a message. `auto` means the model can pick between generating a message or calling one or more tools. `any` or `required` means the model must call one or more tools. Specifying a particular tool via `{\"type\": \"function\", \"function\": {\"name\": \"my_function\"}}` forces the model to call that tool.
+        :param presence_penalty: The `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: The `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 prediction:
-        :param parallel_tool_calls:
+        :param prediction: Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content.
+        :param parallel_tool_calls: Whether to enable parallel function calling during tool use, when enabled the model can call multiple tools in parallel.
         :param prompt_mode: Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used.
         :param safe_prompt: Whether to inject a safety prompt before all conversations.
         :param retries: Override the default retry configuration for this method
@@ -332,6 +337,7 @@ class Chat(BaseSDK):
             stream=stream,
             stop=stop,
             random_seed=random_seed,
+            metadata=metadata,
             messages=utils.get_pydantic_model(messages, List[models.Messages]),
             response_format=utils.get_pydantic_model(
                 response_format, Optional[models.ResponseFormat]
@@ -429,6 +435,7 @@ class Chat(BaseSDK):
             ]
         ] = None,
         random_seed: OptionalNullable[int] = UNSET,
+        metadata: OptionalNullable[Dict[str, Any]] = UNSET,
         response_format: Optional[
             Union[models.ResponseFormat, models.ResponseFormatTypedDict]
         ] = None,
@@ -467,14 +474,15 @@ class Chat(BaseSDK):
         :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 metadata:
+        :param response_format: Specify the format that the model must output. By default it will use `{ \"type\": \"text\" }`. Setting to `{ \"type\": \"json_object\" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ \"type\": \"json_schema\" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide.
+        :param tools: A list of tools the model may call. Use this to provide a list of functions the model may generate JSON inputs for.
+        :param tool_choice: Controls which (if any) tool is called by the model. `none` means the model will not call any tool and instead generates a message. `auto` means the model can pick between generating a message or calling one or more tools. `any` or `required` means the model must call one or more tools. Specifying a particular tool via `{\"type\": \"function\", \"function\": {\"name\": \"my_function\"}}` forces the model to call that tool.
+        :param presence_penalty: The `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: The `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 prediction:
-        :param parallel_tool_calls:
+        :param prediction: Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content.
+        :param parallel_tool_calls: Whether to enable parallel function calling during tool use, when enabled the model can call multiple tools in parallel.
         :param prompt_mode: Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used.
         :param safe_prompt: Whether to inject a safety prompt before all conversations.
         :param retries: Override the default retry configuration for this method
@@ -500,6 +508,7 @@ class Chat(BaseSDK):
             stream=stream,
             stop=stop,
             random_seed=random_seed,
+            metadata=metadata,
             messages=utils.get_pydantic_model(
                 messages, List[models.ChatCompletionStreamRequestMessages]
             ),
@@ -607,6 +616,7 @@ class Chat(BaseSDK):
             ]
         ] = None,
         random_seed: OptionalNullable[int] = UNSET,
+        metadata: OptionalNullable[Dict[str, Any]] = UNSET,
         response_format: Optional[
             Union[models.ResponseFormat, models.ResponseFormatTypedDict]
         ] = None,
@@ -645,14 +655,15 @@ class Chat(BaseSDK):
         :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 metadata:
+        :param response_format: Specify the format that the model must output. By default it will use `{ \"type\": \"text\" }`. Setting to `{ \"type\": \"json_object\" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ \"type\": \"json_schema\" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide.
+        :param tools: A list of tools the model may call. Use this to provide a list of functions the model may generate JSON inputs for.
+        :param tool_choice: Controls which (if any) tool is called by the model. `none` means the model will not call any tool and instead generates a message. `auto` means the model can pick between generating a message or calling one or more tools. `any` or `required` means the model must call one or more tools. Specifying a particular tool via `{\"type\": \"function\", \"function\": {\"name\": \"my_function\"}}` forces the model to call that tool.
+        :param presence_penalty: The `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: The `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 prediction:
-        :param parallel_tool_calls:
+        :param prediction: Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content.
+        :param parallel_tool_calls: Whether to enable parallel function calling during tool use, when enabled the model can call multiple tools in parallel.
         :param prompt_mode: Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used.
         :param safe_prompt: Whether to inject a safety prompt before all conversations.
         :param retries: Override the default retry configuration for this method
@@ -678,6 +689,7 @@ class Chat(BaseSDK):
             stream=stream,
             stop=stop,
             random_seed=random_seed,
+            metadata=metadata,
             messages=utils.get_pydantic_model(
                 messages, List[models.ChatCompletionStreamRequestMessages]
             ),