mistralai 1.4.0__tar.gz → 1.5.1__tar.gz

{mistralai-1.4.0 → mistralai-1.5.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: mistralai
-Version: 1.4.0
+Version: 1.5.1
 Summary: Python Client SDK for the Mistral AI API.
 Author: Mistral
 Requires-Python: >=3.8
@@ -67,6 +67,7 @@ Mistral AI API: Our Chat Completion and Embeddings APIs specification. Create yo
   * [Server Selection](https://github.com/mistralai/client-python/blob/master/#server-selection)
   * [Custom HTTP Client](https://github.com/mistralai/client-python/blob/master/#custom-http-client)
   * [Authentication](https://github.com/mistralai/client-python/blob/master/#authentication)
+  * [Resource Management](https://github.com/mistralai/client-python/blob/master/#resource-management)
   * [Debugging](https://github.com/mistralai/client-python/blob/master/#debugging)
   * [IDE Support](https://github.com/mistralai/client-python/blob/master/#ide-support)
 * [Development](https://github.com/mistralai/client-python/blob/master/#development)
@@ -77,6 +78,11 @@ Mistral AI API: Our Chat Completion and Embeddings APIs specification. Create yo
 <!-- Start SDK Installation [installation] -->
 ## SDK Installation
 
+> [!NOTE]
+> **Python version upgrade policy**
+>
+> Once a Python version reaches its [official end of life date](https://devguide.python.org/versions/), a 3-month grace period is provided for users to upgrade. Following this grace period, the minimum python version supported in the SDK will be updated.
+
 The SDK can be installed with either *pip* or *poetry* package managers.
 
 ### PIP
@@ -461,6 +467,10 @@ The documentation for the GCP SDK is available [here](https://github.com/mistral
 * [archive](https://github.com/mistralai/client-python/blob/master/docs/sdks/models/README.md#archive) - Archive Fine Tuned Model
 * [unarchive](https://github.com/mistralai/client-python/blob/master/docs/sdks/models/README.md#unarchive) - Unarchive Fine Tuned Model
 
+### [ocr](https://github.com/mistralai/client-python/blob/master/docs/sdks/ocr/README.md)
+
+* [process](https://github.com/mistralai/client-python/blob/master/docs/sdks/ocr/README.md#process) - OCR
+
 </details>
 <!-- End Available Resources and Operations [operations] -->
 
@@ -779,6 +789,32 @@ with Mistral(
 ```
 <!-- End Authentication [security] -->
 
+<!-- Start Resource Management [resource-management] -->
+## Resource Management
+
+The `Mistral` class implements the context manager protocol and registers a finalizer function to close the underlying sync and async HTTPX clients it uses under the hood. This will close HTTP connections, release memory and free up other resources held by the SDK. In short-lived Python programs and notebooks that make a few SDK method calls, resource management may not be a concern. However, in longer-lived programs, it is beneficial to create a single SDK instance via a [context manager][context-manager] and reuse it across the application.
+
+[context-manager]: https://docs.python.org/3/reference/datamodel.html#context-managers
+
+```python
+from mistralai import Mistral
+import os
+def main():
+    with Mistral(
+        api_key=os.getenv("MISTRAL_API_KEY", ""),
+    ) as mistral:
+        # Rest of application here...
+
+
+# Or when using async:
+async def amain():
+    async with Mistral(
+        api_key=os.getenv("MISTRAL_API_KEY", ""),
+    ) as mistral:
+        # Rest of application here...
+```
+<!-- End Resource Management [resource-management] -->
+
 <!-- Start Debugging [debug] -->
 ## Debugging
 

{mistralai-1.4.0 → mistralai-1.5.1}/README-PYPI.md

@@ -42,6 +42,7 @@ Mistral AI API: Our Chat Completion and Embeddings APIs specification. Create yo
   * [Server Selection](https://github.com/mistralai/client-python/blob/master/#server-selection)
   * [Custom HTTP Client](https://github.com/mistralai/client-python/blob/master/#custom-http-client)
   * [Authentication](https://github.com/mistralai/client-python/blob/master/#authentication)
+  * [Resource Management](https://github.com/mistralai/client-python/blob/master/#resource-management)
   * [Debugging](https://github.com/mistralai/client-python/blob/master/#debugging)
   * [IDE Support](https://github.com/mistralai/client-python/blob/master/#ide-support)
 * [Development](https://github.com/mistralai/client-python/blob/master/#development)
@@ -52,6 +53,11 @@ Mistral AI API: Our Chat Completion and Embeddings APIs specification. Create yo
 <!-- Start SDK Installation [installation] -->
 ## SDK Installation
 
+> [!NOTE]
+> **Python version upgrade policy**
+>
+> Once a Python version reaches its [official end of life date](https://devguide.python.org/versions/), a 3-month grace period is provided for users to upgrade. Following this grace period, the minimum python version supported in the SDK will be updated.
+
 The SDK can be installed with either *pip* or *poetry* package managers.
 
 ### PIP
@@ -436,6 +442,10 @@ The documentation for the GCP SDK is available [here](https://github.com/mistral
 * [archive](https://github.com/mistralai/client-python/blob/master/docs/sdks/models/README.md#archive) - Archive Fine Tuned Model
 * [unarchive](https://github.com/mistralai/client-python/blob/master/docs/sdks/models/README.md#unarchive) - Unarchive Fine Tuned Model
 
+### [ocr](https://github.com/mistralai/client-python/blob/master/docs/sdks/ocr/README.md)
+
+* [process](https://github.com/mistralai/client-python/blob/master/docs/sdks/ocr/README.md#process) - OCR
+
 </details>
 <!-- End Available Resources and Operations [operations] -->
 
@@ -754,6 +764,32 @@ with Mistral(
 ```
 <!-- End Authentication [security] -->
 
+<!-- Start Resource Management [resource-management] -->
+## Resource Management
+
+The `Mistral` class implements the context manager protocol and registers a finalizer function to close the underlying sync and async HTTPX clients it uses under the hood. This will close HTTP connections, release memory and free up other resources held by the SDK. In short-lived Python programs and notebooks that make a few SDK method calls, resource management may not be a concern. However, in longer-lived programs, it is beneficial to create a single SDK instance via a [context manager][context-manager] and reuse it across the application.
+
+[context-manager]: https://docs.python.org/3/reference/datamodel.html#context-managers
+
+```python
+from mistralai import Mistral
+import os
+def main():
+    with Mistral(
+        api_key=os.getenv("MISTRAL_API_KEY", ""),
+    ) as mistral:
+        # Rest of application here...
+
+
+# Or when using async:
+async def amain():
+    async with Mistral(
+        api_key=os.getenv("MISTRAL_API_KEY", ""),
+    ) as mistral:
+        # Rest of application here...
+```
+<!-- End Resource Management [resource-management] -->
+
 <!-- Start Debugging [debug] -->
 ## Debugging
 

{mistralai-1.4.0 → mistralai-1.5.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "mistralai"
-version = "1.4.0"
+version = "1.5.1"
 description = "Python Client SDK for the Mistral AI API."
 authors = ["Mistral"]
 readme = "README-PYPI.md"

{mistralai-1.4.0 → mistralai-1.5.1}/src/mistralai/_version.py

@@ -3,10 +3,10 @@
 import importlib.metadata
 
 __title__: str = "mistralai"
-__version__: str = "1.4.0"
+__version__: str = "1.5.1"
 __openapi_doc_version__: str = "0.0.2"
-__gen_version__: str = "2.493.32"
-__user_agent__: str = "speakeasy-sdk/python 1.4.0 2.493.32 0.0.2 mistralai"
+__gen_version__: str = "2.497.0"
+__user_agent__: str = "speakeasy-sdk/python 1.5.1 2.497.0 0.0.2 mistralai"
 
 try:
     if __package__ is not None:

{mistralai-1.4.0 → mistralai-1.5.1}/src/mistralai/chat.py

@@ -3,18 +3,100 @@
 from .basesdk import BaseSDK
 from mistralai import models, utils
 from mistralai._hooks import HookContext
-from mistralai.types import Nullable, OptionalNullable, UNSET
+from mistralai.types import OptionalNullable, UNSET
 from mistralai.utils import eventstreaming, get_security_from_env
 from typing import Any, List, Mapping, Optional, Union
 
+# region imports
+from typing import Type
+from mistralai.extra import (
+    convert_to_parsed_chat_completion_response,
+    response_format_from_pydantic_model,
+    CustomPydanticModel,
+    ParsedChatCompletionResponse,
+)
+# endregion imports
+
 
 class Chat(BaseSDK):
     r"""Chat Completion API."""
 
+    # region sdk-class-body
+    # Custom .parse methods for the Structure Outputs Feature.
+
+    def parse(
+        self, response_format: Type[CustomPydanticModel], **kwargs: Any
+    ) -> ParsedChatCompletionResponse[CustomPydanticModel]:
+        """
+        Parse the response using the provided response format.
+        :param Type[CustomPydanticModel] response_format: The Pydantic model to parse the response into
+        :param Any **kwargs Additional keyword arguments to pass to the .complete method
+        :return: The parsed response
+        """
+        # Convert the input Pydantic Model to a strict JSON ready to be passed to chat.complete
+        json_response_format = response_format_from_pydantic_model(response_format)
+        # Run the inference
+        response = self.complete(**kwargs, response_format=json_response_format)
+        # Parse response back to the input pydantic model
+        parsed_response = convert_to_parsed_chat_completion_response(
+            response, response_format
+        )
+        return parsed_response
+
+    async def parse_async(
+        self, response_format: Type[CustomPydanticModel], **kwargs
+    ) -> ParsedChatCompletionResponse[CustomPydanticModel]:
+        """
+        Asynchronously parse the response using the provided response format.
+        :param Type[CustomPydanticModel] response_format: The Pydantic model to parse the response into
+        :param Any **kwargs Additional keyword arguments to pass to the .complete method
+        :return: The parsed response
+        """
+        json_response_format = response_format_from_pydantic_model(response_format)
+        response = await self.complete_async(  # pylint: disable=E1125
+            **kwargs, response_format=json_response_format
+        )
+        parsed_response = convert_to_parsed_chat_completion_response(
+            response, response_format
+        )
+        return parsed_response
+
+    def parse_stream(
+        self, response_format: Type[CustomPydanticModel], **kwargs
+    ) -> eventstreaming.EventStream[models.CompletionEvent]:
+        """
+        Parse the response using the provided response format.
+        For now the response will be in JSON format not in the input Pydantic model.
+        :param Type[CustomPydanticModel] response_format: The Pydantic model to parse the response into
+        :param Any **kwargs Additional keyword arguments to pass to the .stream method
+        :return: The JSON parsed response
+        """
+        json_response_format = response_format_from_pydantic_model(response_format)
+        response = self.stream(**kwargs, response_format=json_response_format)
+        return response
+
+    async def parse_stream_async(
+        self, response_format: Type[CustomPydanticModel], **kwargs
+    ) -> eventstreaming.EventStreamAsync[models.CompletionEvent]:
+        """
+        Asynchronously parse the response using the provided response format.
+        For now the response will be in JSON format not in the input Pydantic model.
+        :param Type[CustomPydanticModel] response_format: The Pydantic model to parse the response into
+        :param Any **kwargs Additional keyword arguments to pass to the .stream method
+        :return: The JSON parsed response
+        """
+        json_response_format = response_format_from_pydantic_model(response_format)
+        response = await self.stream_async(  # pylint: disable=E1125
+            **kwargs, response_format=json_response_format
+        )
+        return response
+
+    # endregion sdk-class-body
+
     def complete(
         self,
         *,
-        model: Nullable[str],
+        model: str,
         messages: Union[List[models.Messages], List[models.MessagesTypedDict]],
         temperature: OptionalNullable[float] = UNSET,
         top_p: Optional[float] = None,
@@ -171,7 +253,7 @@ class Chat(BaseSDK):
     async def complete_async(
         self,
         *,
-        model: Nullable[str],
+        model: str,
         messages: Union[List[models.Messages], List[models.MessagesTypedDict]],
         temperature: OptionalNullable[float] = UNSET,
         top_p: Optional[float] = None,
@@ -328,7 +410,7 @@ class Chat(BaseSDK):
     def stream(
         self,
         *,
-        model: Nullable[str],
+        model: str,
         messages: Union[
             List[models.ChatCompletionStreamRequestMessages],
             List[models.ChatCompletionStreamRequestMessagesTypedDict],
@@ -503,7 +585,7 @@ class Chat(BaseSDK):
     async def stream_async(
         self,
         *,
-        model: Nullable[str],
+        model: str,
         messages: Union[
             List[models.ChatCompletionStreamRequestMessages],
             List[models.ChatCompletionStreamRequestMessagesTypedDict],

{mistralai-1.4.0 → mistralai-1.5.1}/src/mistralai/classifiers.py

@@ -3,7 +3,7 @@
 from .basesdk import BaseSDK
 from mistralai import models, utils
 from mistralai._hooks import HookContext
-from mistralai.types import Nullable, OptionalNullable, UNSET
+from mistralai.types import OptionalNullable, UNSET
 from mistralai.utils import get_security_from_env
 from typing import Any, Mapping, Optional, Union
 
@@ -14,11 +14,11 @@ class Classifiers(BaseSDK):
     def moderate(
         self,
         *,
+        model: str,
         inputs: Union[
             models.ClassificationRequestInputs,
             models.ClassificationRequestInputsTypedDict,
         ],
-        model: OptionalNullable[str] = UNSET,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -26,8 +26,8 @@ class Classifiers(BaseSDK):
     ) -> models.ClassificationResponse:
         r"""Moderations
 
+        :param model: ID of the model to use.
         :param inputs: Text to classify.
-        :param model:
         :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
@@ -42,8 +42,8 @@ class Classifiers(BaseSDK):
             base_url = server_url
 
         request = models.ClassificationRequest(
-            inputs=inputs,
             model=model,
+            inputs=inputs,
         )
 
         req = self._build_request(
@@ -115,11 +115,11 @@ class Classifiers(BaseSDK):
     async def moderate_async(
         self,
         *,
+        model: str,
         inputs: Union[
             models.ClassificationRequestInputs,
             models.ClassificationRequestInputsTypedDict,
         ],
-        model: OptionalNullable[str] = UNSET,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -127,8 +127,8 @@ class Classifiers(BaseSDK):
     ) -> models.ClassificationResponse:
         r"""Moderations
 
+        :param model: ID of the model to use.
         :param inputs: Text to classify.
-        :param model:
         :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
@@ -143,8 +143,8 @@ class Classifiers(BaseSDK):
             base_url = server_url
 
         request = models.ClassificationRequest(
-            inputs=inputs,
             model=model,
+            inputs=inputs,
         )
 
         req = self._build_request_async(
@@ -216,11 +216,12 @@ class Classifiers(BaseSDK):
     def moderate_chat(
         self,
         *,
+        model: str,
         inputs: Union[
-            models.ChatClassificationRequestInputs,
-            models.ChatClassificationRequestInputsTypedDict,
+            models.ChatModerationRequestInputs,
+            models.ChatModerationRequestInputsTypedDict,
         ],
-        model: Nullable[str],
+        truncate_for_context_length: Optional[bool] = False,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -228,8 +229,9 @@ class Classifiers(BaseSDK):
     ) -> models.ClassificationResponse:
         r"""Moderations Chat
 
-        :param inputs: Chat to classify
         :param model:
+        :param inputs: Chat to classify
+        :param truncate_for_context_length:
         :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
@@ -243,11 +245,10 @@ class Classifiers(BaseSDK):
         if server_url is not None:
             base_url = server_url
 
-        request = models.ChatClassificationRequest(
-            inputs=utils.get_pydantic_model(
-                inputs, models.ChatClassificationRequestInputs
-            ),
+        request = models.ChatModerationRequest(
             model=model,
+            inputs=utils.get_pydantic_model(inputs, models.ChatModerationRequestInputs),
+            truncate_for_context_length=truncate_for_context_length,
         )
 
         req = self._build_request(
@@ -264,7 +265,7 @@ class Classifiers(BaseSDK):
             http_headers=http_headers,
             security=self.sdk_configuration.security,
             get_serialized_body=lambda: utils.serialize_request_body(
-                request, False, False, "json", models.ChatClassificationRequest
+                request, False, False, "json", models.ChatModerationRequest
             ),
             timeout_ms=timeout_ms,
         )
@@ -319,11 +320,12 @@ class Classifiers(BaseSDK):
     async def moderate_chat_async(
         self,
         *,
+        model: str,
         inputs: Union[
-            models.ChatClassificationRequestInputs,
-            models.ChatClassificationRequestInputsTypedDict,
+            models.ChatModerationRequestInputs,
+            models.ChatModerationRequestInputsTypedDict,
         ],
-        model: Nullable[str],
+        truncate_for_context_length: Optional[bool] = False,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -331,8 +333,9 @@ class Classifiers(BaseSDK):
     ) -> models.ClassificationResponse:
         r"""Moderations Chat
 
-        :param inputs: Chat to classify
         :param model:
+        :param inputs: Chat to classify
+        :param truncate_for_context_length:
         :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
@@ -346,11 +349,10 @@ class Classifiers(BaseSDK):
         if server_url is not None:
             base_url = server_url
 
-        request = models.ChatClassificationRequest(
-            inputs=utils.get_pydantic_model(
-                inputs, models.ChatClassificationRequestInputs
-            ),
+        request = models.ChatModerationRequest(
             model=model,
+            inputs=utils.get_pydantic_model(inputs, models.ChatModerationRequestInputs),
+            truncate_for_context_length=truncate_for_context_length,
         )
 
         req = self._build_request_async(
@@ -367,7 +369,7 @@ class Classifiers(BaseSDK):
             http_headers=http_headers,
             security=self.sdk_configuration.security,
             get_serialized_body=lambda: utils.serialize_request_body(
-                request, False, False, "json", models.ChatClassificationRequest
+                request, False, False, "json", models.ChatModerationRequest
             ),
             timeout_ms=timeout_ms,
         )

{mistralai-1.4.0 → mistralai-1.5.1}/src/mistralai/embeddings.py

@@ -16,7 +16,6 @@ class Embeddings(BaseSDK):
         *,
         inputs: Union[models.Inputs, models.InputsTypedDict],
         model: Optional[str] = "mistral-embed",
-        encoding_format: OptionalNullable[str] = UNSET,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -28,7 +27,6 @@ class Embeddings(BaseSDK):
 
         :param inputs: Text to embed.
         :param model: ID of the model to use.
-        :param encoding_format: The format to return the embeddings in.
         :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
@@ -43,9 +41,8 @@ class Embeddings(BaseSDK):
             base_url = server_url
 
         request = models.EmbeddingRequest(
-            inputs=inputs,
             model=model,
-            encoding_format=encoding_format,
+            inputs=inputs,
         )
 
         req = self._build_request(
@@ -119,7 +116,6 @@ class Embeddings(BaseSDK):
         *,
         inputs: Union[models.Inputs, models.InputsTypedDict],
         model: Optional[str] = "mistral-embed",
-        encoding_format: OptionalNullable[str] = UNSET,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
@@ -131,7 +127,6 @@ class Embeddings(BaseSDK):
 
         :param inputs: Text to embed.
         :param model: ID of the model to use.
-        :param encoding_format: The format to return the embeddings in.
         :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
@@ -146,9 +141,8 @@ class Embeddings(BaseSDK):
             base_url = server_url
 
         request = models.EmbeddingRequest(
-            inputs=inputs,
             model=model,
-            encoding_format=encoding_format,
+            inputs=inputs,
         )
 
         req = self._build_request_async(

mistralai-1.5.1/src/mistralai/extra/README.md

@@ -0,0 +1,56 @@
+## Context
+
+The extra package contains the custom logic which is too complex to be generated by Speakeasy from the OpenAPI specs. It was introduced to add the Structured Outputs feature.
+
+## Development / Contributing
+
+To add custom code in the SDK, you need to use [Speakeasy custom code regions](https://www.speakeasy.com/docs/customize/code/code-regions/overview) as below.
+
+### Runbook of SDK customization
+
+1. Add the code you want to import in the `src/mistralai/extra/` package. To have it importable from the SDK, you need to add it in the `__init__.py` file:
+```python
+from .my_custom_file import my_custom_function
+
+__all__ = ["my_custom_function"]
+```
+
+2. Add a new custom code region in the SDK files, e.g in `src/mistralai/chat.py`:
+```python
+# region imports
+from typing import Type
+from mistralai.extra import my_custom_function
+# endregion imports
+
+class Chat(BaseSDK):
+    r"""Chat Completion API."""
+
+    # region sdk-class-body
+    def my_custom_method(self, param: str) -> Type[some_type]:
+        output = my_custom_function(param1)
+        return output
+    # endregion sdk-class-body
+```
+
+3. Now build the SDK with the custom code:
+```bash
+rm -rf dist; poetry build; python3 -m pip install ~/client-python/dist/mistralai-1.4.1-py3-none-any.whl --force-reinstall
+```
+
+4. And now you should be able to call the custom method:
+```python
+import os
+from mistralai import Mistral
+
+api_key = os.environ["MISTRAL_API_KEY"]
+client = Mistral(api_key=api_key)
+
+client.chat.my_custom_method(param="test")
+```
+
+### Run the unit tests
+
+To run the unit tests for the `extra` package, you can run the following command from the root of the repository:
+```bash
+python3.12 -m unittest discover -s src/mistralai/extra/tests -t src
+```

mistralai-1.5.1/src/mistralai/extra/__init__.py

@@ -0,0 +1,5 @@
+from .struct_chat import ParsedChatCompletionResponse, convert_to_parsed_chat_completion_response
+from .utils import response_format_from_pydantic_model
+from .utils.response_format import CustomPydanticModel
+
+__all__ = ["convert_to_parsed_chat_completion_response", "response_format_from_pydantic_model", "CustomPydanticModel", "ParsedChatCompletionResponse"]

mistralai-1.5.1/src/mistralai/extra/struct_chat.py

@@ -0,0 +1,41 @@
+from ..models import ChatCompletionResponse, ChatCompletionChoice, AssistantMessage
+from .utils.response_format import  CustomPydanticModel, pydantic_model_from_json
+from typing import List, Optional, Type, Generic
+from pydantic import BaseModel
+import json
+
+class ParsedAssistantMessage(AssistantMessage, Generic[CustomPydanticModel]):
+    parsed: Optional[CustomPydanticModel]
+
+class ParsedChatCompletionChoice(ChatCompletionChoice, Generic[CustomPydanticModel]):
+    message: Optional[ParsedAssistantMessage[CustomPydanticModel]]  # type: ignore
+
+class ParsedChatCompletionResponse(ChatCompletionResponse, Generic[CustomPydanticModel]):
+    choices: Optional[List[ParsedChatCompletionChoice[CustomPydanticModel]]] # type: ignore
+
+def convert_to_parsed_chat_completion_response(response: ChatCompletionResponse, response_format: Type[BaseModel]) -> ParsedChatCompletionResponse:
+    parsed_choices = []
+
+    if response.choices:
+        for choice in response.choices:
+            if choice.message:
+                parsed_message: ParsedAssistantMessage = ParsedAssistantMessage(
+                    **choice.message.model_dump(),
+                    parsed=None
+                )
+                if isinstance(parsed_message.content, str):
+                    parsed_message.parsed = pydantic_model_from_json(json.loads(parsed_message.content), response_format)
+                elif parsed_message.content is None:
+                    parsed_message.parsed = None
+                else:
+                    raise TypeError(f"Unexpected type for message.content: {type(parsed_message.content)}")
+                choice_dict = choice.model_dump()
+                choice_dict["message"] = parsed_message
+                parsed_choice: ParsedChatCompletionChoice = ParsedChatCompletionChoice(**choice_dict)
+                parsed_choices.append(parsed_choice)
+            else:
+                parsed_choice = ParsedChatCompletionChoice(**choice.model_dump())
+                parsed_choices.append(parsed_choice)
+    response_dict = response.model_dump()
+    response_dict["choices"] = parsed_choices
+    return ParsedChatCompletionResponse(**response_dict)