mistralai 1.10.0__py3-none-any.whl → 1.10.1__py3-none-any.whl

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)

mistralai_gcp/fim.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, Mapping, Optional, Union
 
 
@@ -38,7 +39,7 @@ class Fim(BaseSDK):
 
         Mistral AI provides the ability to stream responses back to a client in order to allow partial results for certain requests. Tokens will be sent as data-only server-sent 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 model: ID of the model to use. Only compatible for now with:   - `codestral-2405`   - `codestral-latest`
+        :param model: ID of the model with FIM to use.
         :param prompt: The text/code to complete.
         :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.
@@ -105,6 +106,7 @@ class Fim(BaseSDK):
 
         http_res = self.do_request(
             hook_ctx=HookContext(
+                config=self.sdk_configuration,
                 base_url=base_url or "",
                 operation_id="stream_fim",
                 oauth2_scopes=[],
@@ -122,32 +124,23 @@ class Fim(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,
@@ -176,7 +169,7 @@ class Fim(BaseSDK):
 
         Mistral AI provides the ability to stream responses back to a client in order to allow partial results for certain requests. Tokens will be sent as data-only server-sent 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 model: ID of the model to use. Only compatible for now with:   - `codestral-2405`   - `codestral-latest`
+        :param model: ID of the model with FIM to use.
         :param prompt: The text/code to complete.
         :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.
@@ -243,6 +236,7 @@ class Fim(BaseSDK):
 
         http_res = await self.do_request_async(
             hook_ctx=HookContext(
+                config=self.sdk_configuration,
                 base_url=base_url or "",
                 operation_id="stream_fim",
                 oauth2_scopes=[],
@@ -260,32 +254,23 @@ class Fim(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,
@@ -314,7 +299,7 @@ class Fim(BaseSDK):
 
         FIM completion.
 
-        :param model: ID of the model to use. Only compatible for now with:   - `codestral-2405`   - `codestral-latest`
+        :param model: ID of the model with FIM to use.
         :param prompt: The text/code to complete.
         :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.
@@ -381,6 +366,7 @@ class Fim(BaseSDK):
 
         http_res = self.do_request(
             hook_ctx=HookContext(
+                config=self.sdk_configuration,
                 base_url=base_url or "",
                 operation_id="fim_completion_v1_fim_completions_post",
                 oauth2_scopes=[],
@@ -393,33 +379,22 @@ class Fim(BaseSDK):
 
         response_data: Any = None
         if utils.match_response(http_res, "200", "application/json"):
-            return utils.unmarshal_json(
-                http_res.text, Optional[models.FIMCompletionResponse]
+            return unmarshal_json_response(
+                Optional[models.FIMCompletionResponse], 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,
@@ -448,7 +423,7 @@ class Fim(BaseSDK):
 
         FIM completion.
 
-        :param model: ID of the model to use. Only compatible for now with:   - `codestral-2405`   - `codestral-latest`
+        :param model: ID of the model with FIM to use.
         :param prompt: The text/code to complete.
         :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.
@@ -515,6 +490,7 @@ class Fim(BaseSDK):
 
         http_res = await self.do_request_async(
             hook_ctx=HookContext(
+                config=self.sdk_configuration,
                 base_url=base_url or "",
                 operation_id="fim_completion_v1_fim_completions_post",
                 oauth2_scopes=[],
@@ -527,30 +503,19 @@ class Fim(BaseSDK):
 
         response_data: Any = None
         if utils.match_response(http_res, "200", "application/json"):
-            return utils.unmarshal_json(
-                http_res.text, Optional[models.FIMCompletionResponse]
+            return unmarshal_json_response(
+                Optional[models.FIMCompletionResponse], 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)

mistralai_gcp/httpclient.py

@@ -2,7 +2,6 @@
 
 # pyright: reportReturnType = false
 import asyncio
-from concurrent.futures import ThreadPoolExecutor
 from typing_extensions import Protocol, runtime_checkable
 import httpx
 from typing import Any, Optional, Union
@@ -116,21 +115,12 @@ def close_clients(
             pass
 
     if async_client is not None and not async_client_supplied:
-        is_async = False
         try:
-            asyncio.get_running_loop()
-            is_async = True
+            loop = asyncio.get_running_loop()
+            asyncio.run_coroutine_threadsafe(async_client.aclose(), loop)
         except RuntimeError:
-            pass
-
-        try:
-            # If this function is called in an async loop then start another
-            # loop in a separate thread to close the async http client.
-            if is_async:
-                with ThreadPoolExecutor(max_workers=1) as executor:
-                    future = executor.submit(asyncio.run, async_client.aclose())
-                    future.result()
-            else:
+            try:
                 asyncio.run(async_client.aclose())
-        except Exception:
-            pass
+            except RuntimeError:
+                # best effort
+                pass