mistralai 1.9.11__py3-none-any.whl → 1.10.1__py3-none-any.whl

mistralai_gcp/basesdk.py

@@ -15,9 +15,19 @@ from urllib.parse import parse_qs, urlparse
 
 class BaseSDK:
     sdk_configuration: SDKConfiguration
+    parent_ref: Optional[object] = None
+    """
+    Reference to the root SDK instance, if any. This will prevent it from
+    being garbage collected while there are active streams.
+    """
 
-    def __init__(self, sdk_config: SDKConfiguration) -> None:
+    def __init__(
+        self,
+        sdk_config: SDKConfiguration,
+        parent_ref: Optional[object] = None,
+    ) -> None:
         self.sdk_configuration = sdk_config
+        self.parent_ref = parent_ref
 
     def _get_url(self, base_url, url_variables):
         sdk_url, sdk_variables = self.sdk_configuration.get_server_details()
@@ -218,12 +228,12 @@ class BaseSDK:
         client = self.sdk_configuration.client
         logger = self.sdk_configuration.debug_logger
 
+        hooks = self.sdk_configuration.__dict__["_hooks"]
+
         def do():
             http_res = None
             try:
-                req = self.sdk_configuration.get_hooks().before_request(
-                    BeforeRequestContext(hook_ctx), request
-                )
+                req = hooks.before_request(BeforeRequestContext(hook_ctx), request)
                 logger.debug(
                     "Request:\nMethod: %s\nURL: %s\nHeaders: %s\nBody: %s",
                     req.method,
@@ -237,16 +247,14 @@ class BaseSDK:
 
                 http_res = client.send(req, stream=stream)
             except Exception as e:
-                _, e = self.sdk_configuration.get_hooks().after_error(
-                    AfterErrorContext(hook_ctx), None, e
-                )
+                _, e = hooks.after_error(AfterErrorContext(hook_ctx), None, e)
                 if e is not None:
                     logger.debug("Request Exception", exc_info=True)
                     raise e
 
             if http_res is None:
                 logger.debug("Raising no response SDK error")
-                raise models.SDKError("No response received")
+                raise models.NoResponseError("No response received")
 
             logger.debug(
                 "Response:\nStatus Code: %s\nURL: %s\nHeaders: %s\nBody: %s",
@@ -257,7 +265,7 @@ class BaseSDK:
             )
 
             if utils.match_status_codes(error_status_codes, http_res.status_code):
-                result, err = self.sdk_configuration.get_hooks().after_error(
+                result, err = hooks.after_error(
                     AfterErrorContext(hook_ctx), http_res, None
                 )
                 if err is not None:
@@ -267,7 +275,7 @@ class BaseSDK:
                     http_res = result
                 else:
                     logger.debug("Raising unexpected SDK error")
-                    raise models.SDKError("Unexpected error occurred")
+                    raise models.SDKError("Unexpected error occurred", http_res)
 
             return http_res
 
@@ -277,9 +285,7 @@ class BaseSDK:
             http_res = do()
 
         if not utils.match_status_codes(error_status_codes, http_res.status_code):
-            http_res = self.sdk_configuration.get_hooks().after_success(
-                AfterSuccessContext(hook_ctx), http_res
-            )
+            http_res = hooks.after_success(AfterSuccessContext(hook_ctx), http_res)
 
         return http_res
 
@@ -294,12 +300,12 @@ class BaseSDK:
         client = self.sdk_configuration.async_client
         logger = self.sdk_configuration.debug_logger
 
+        hooks = self.sdk_configuration.__dict__["_hooks"]
+
         async def do():
             http_res = None
             try:
-                req = self.sdk_configuration.get_hooks().before_request(
-                    BeforeRequestContext(hook_ctx), request
-                )
+                req = hooks.before_request(BeforeRequestContext(hook_ctx), request)
                 logger.debug(
                     "Request:\nMethod: %s\nURL: %s\nHeaders: %s\nBody: %s",
                     req.method,
@@ -313,16 +319,14 @@ class BaseSDK:
 
                 http_res = await client.send(req, stream=stream)
             except Exception as e:
-                _, e = self.sdk_configuration.get_hooks().after_error(
-                    AfterErrorContext(hook_ctx), None, e
-                )
+                _, e = hooks.after_error(AfterErrorContext(hook_ctx), None, e)
                 if e is not None:
                     logger.debug("Request Exception", exc_info=True)
                     raise e
 
             if http_res is None:
                 logger.debug("Raising no response SDK error")
-                raise models.SDKError("No response received")
+                raise models.NoResponseError("No response received")
 
             logger.debug(
                 "Response:\nStatus Code: %s\nURL: %s\nHeaders: %s\nBody: %s",
@@ -333,7 +337,7 @@ class BaseSDK:
             )
 
             if utils.match_status_codes(error_status_codes, http_res.status_code):
-                result, err = self.sdk_configuration.get_hooks().after_error(
+                result, err = hooks.after_error(
                     AfterErrorContext(hook_ctx), http_res, None
                 )
                 if err is not None:
@@ -343,7 +347,7 @@ class BaseSDK:
                     http_res = result
                 else:
                     logger.debug("Raising unexpected SDK error")
-                    raise models.SDKError("Unexpected error occurred")
+                    raise models.SDKError("Unexpected error occurred", http_res)
 
             return http_res
 
@@ -355,8 +359,6 @@ class BaseSDK:
             http_res = await do()
 
         if not utils.match_status_codes(error_status_codes, http_res.status_code):
-            http_res = self.sdk_configuration.get_hooks().after_success(
-                AfterSuccessContext(hook_ctx), http_res
-            )
+            http_res = hooks.after_success(AfterSuccessContext(hook_ctx), http_res)
 
         return http_res

mistralai_gcp/chat.py

@@ -5,6 +5,7 @@ from mistralai_gcp import models, utils
 from mistralai_gcp._hooks import HookContext
 from mistralai_gcp.types import OptionalNullable, UNSET
 from mistralai_gcp.utils import eventstreaming
+from mistralai_gcp.utils.unmarshal_json_response import unmarshal_json_response
 from typing import Any, List, Mapping, Optional, Union
 
 
@@ -41,6 +42,7 @@ class Chat(BaseSDK):
             Union[models.Prediction, models.PredictionTypedDict]
         ] = None,
         parallel_tool_calls: Optional[bool] = None,
+        prompt_mode: OptionalNullable[models.MistralPromptMode] = UNSET,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -58,14 +60,15 @@ 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 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 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
@@ -104,6 +107,7 @@ class Chat(BaseSDK):
                 prediction, Optional[models.Prediction]
             ),
             parallel_tool_calls=parallel_tool_calls,
+            prompt_mode=prompt_mode,
         )
 
         req = self._build_request(
@@ -135,6 +139,7 @@ class Chat(BaseSDK):
 
         http_res = self.do_request(
             hook_ctx=HookContext(
+                config=self.sdk_configuration,
                 base_url=base_url or "",
                 operation_id="stream_chat",
                 oauth2_scopes=[],
@@ -152,32 +157,23 @@ class Chat(BaseSDK):
                 http_res,
                 lambda raw: utils.unmarshal_json(raw, models.CompletionEvent),
                 sentinel="[DONE]",
+                client_ref=self,
             )
         if utils.match_response(http_res, "422", "application/json"):
             http_res_text = utils.stream_to_text(http_res)
-            response_data = utils.unmarshal_json(
-                http_res_text, models.HTTPValidationErrorData
+            response_data = unmarshal_json_response(
+                models.HTTPValidationErrorData, http_res, http_res_text
             )
-            raise models.HTTPValidationError(data=response_data)
+            raise models.HTTPValidationError(response_data, http_res, http_res_text)
         if utils.match_response(http_res, "4XX", "*"):
             http_res_text = utils.stream_to_text(http_res)
-            raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res_text, http_res
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
         if utils.match_response(http_res, "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
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
 
-        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,
-        )
+        raise models.SDKError("Unexpected response received", http_res, http_res_text)
 
     async def stream_async(
         self,
@@ -209,6 +205,7 @@ class Chat(BaseSDK):
             Union[models.Prediction, models.PredictionTypedDict]
         ] = None,
         parallel_tool_calls: Optional[bool] = None,
+        prompt_mode: OptionalNullable[models.MistralPromptMode] = UNSET,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -226,14 +223,15 @@ 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 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 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
@@ -272,6 +270,7 @@ class Chat(BaseSDK):
                 prediction, Optional[models.Prediction]
             ),
             parallel_tool_calls=parallel_tool_calls,
+            prompt_mode=prompt_mode,
         )
 
         req = self._build_request_async(
@@ -303,6 +302,7 @@ class Chat(BaseSDK):
 
         http_res = await self.do_request_async(
             hook_ctx=HookContext(
+                config=self.sdk_configuration,
                 base_url=base_url or "",
                 operation_id="stream_chat",
                 oauth2_scopes=[],
@@ -320,32 +320,23 @@ class Chat(BaseSDK):
                 http_res,
                 lambda raw: utils.unmarshal_json(raw, models.CompletionEvent),
                 sentinel="[DONE]",
+                client_ref=self,
             )
         if utils.match_response(http_res, "422", "application/json"):
             http_res_text = await utils.stream_to_text_async(http_res)
-            response_data = utils.unmarshal_json(
-                http_res_text, models.HTTPValidationErrorData
+            response_data = unmarshal_json_response(
+                models.HTTPValidationErrorData, http_res, http_res_text
             )
-            raise models.HTTPValidationError(data=response_data)
+            raise models.HTTPValidationError(response_data, http_res, http_res_text)
         if utils.match_response(http_res, "4XX", "*"):
             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
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
         if utils.match_response(http_res, "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
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
 
-        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,
-        )
+        raise models.SDKError("Unexpected response received", http_res, http_res_text)
 
     def complete(
         self,
@@ -385,6 +376,7 @@ class Chat(BaseSDK):
             Union[models.Prediction, models.PredictionTypedDict]
         ] = None,
         parallel_tool_calls: Optional[bool] = None,
+        prompt_mode: OptionalNullable[models.MistralPromptMode] = UNSET,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -400,14 +392,15 @@ 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 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 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
@@ -448,6 +441,7 @@ class Chat(BaseSDK):
                 prediction, Optional[models.Prediction]
             ),
             parallel_tool_calls=parallel_tool_calls,
+            prompt_mode=prompt_mode,
         )
 
         req = self._build_request(
@@ -479,6 +473,7 @@ class Chat(BaseSDK):
 
         http_res = self.do_request(
             hook_ctx=HookContext(
+                config=self.sdk_configuration,
                 base_url=base_url or "",
                 operation_id="chat_completion_v1_chat_completions_post",
                 oauth2_scopes=[],
@@ -491,33 +486,22 @@ class Chat(BaseSDK):
 
         response_data: Any = None
         if utils.match_response(http_res, "200", "application/json"):
-            return utils.unmarshal_json(
-                http_res.text, Optional[models.ChatCompletionResponse]
+            return unmarshal_json_response(
+                Optional[models.ChatCompletionResponse], http_res
             )
         if utils.match_response(http_res, "422", "application/json"):
-            response_data = utils.unmarshal_json(
-                http_res.text, models.HTTPValidationErrorData
+            response_data = unmarshal_json_response(
+                models.HTTPValidationErrorData, http_res
             )
-            raise models.HTTPValidationError(data=response_data)
+            raise models.HTTPValidationError(response_data, http_res)
         if utils.match_response(http_res, "4XX", "*"):
             http_res_text = utils.stream_to_text(http_res)
-            raise models.SDKError(
-                "API error occurred", http_res.status_code, http_res_text, http_res
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
         if utils.match_response(http_res, "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
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
 
-        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,
-        )
+        raise models.SDKError("Unexpected response received", http_res)
 
     async def complete_async(
         self,
@@ -557,6 +541,7 @@ class Chat(BaseSDK):
             Union[models.Prediction, models.PredictionTypedDict]
         ] = None,
         parallel_tool_calls: Optional[bool] = None,
+        prompt_mode: OptionalNullable[models.MistralPromptMode] = UNSET,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -572,14 +557,15 @@ 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 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 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
@@ -620,6 +606,7 @@ class Chat(BaseSDK):
                 prediction, Optional[models.Prediction]
             ),
             parallel_tool_calls=parallel_tool_calls,
+            prompt_mode=prompt_mode,
         )
 
         req = self._build_request_async(
@@ -651,6 +638,7 @@ class Chat(BaseSDK):
 
         http_res = await self.do_request_async(
             hook_ctx=HookContext(
+                config=self.sdk_configuration,
                 base_url=base_url or "",
                 operation_id="chat_completion_v1_chat_completions_post",
                 oauth2_scopes=[],
@@ -663,30 +651,19 @@ class Chat(BaseSDK):
 
         response_data: Any = None
         if utils.match_response(http_res, "200", "application/json"):
-            return utils.unmarshal_json(
-                http_res.text, Optional[models.ChatCompletionResponse]
+            return unmarshal_json_response(
+                Optional[models.ChatCompletionResponse], http_res
             )
         if utils.match_response(http_res, "422", "application/json"):
-            response_data = utils.unmarshal_json(
-                http_res.text, models.HTTPValidationErrorData
+            response_data = unmarshal_json_response(
+                models.HTTPValidationErrorData, http_res
             )
-            raise models.HTTPValidationError(data=response_data)
+            raise models.HTTPValidationError(response_data, http_res)
         if utils.match_response(http_res, "4XX", "*"):
             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
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
         if utils.match_response(http_res, "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
-            )
+            raise models.SDKError("API error occurred", http_res, http_res_text)
 
-        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,
-        )
+        raise models.SDKError("Unexpected response received", http_res)