mistralai 1.3.1__tar.gz → 1.5.0__tar.gz

{mistralai-1.3.1 → mistralai-1.5.0}/PKG-INFO

@@ -1,8 +1,7 @@
-Metadata-Version: 2.1
+Metadata-Version: 2.3
 Name: mistralai
-Version: 1.3.1
+Version: 1.5.0
 Summary: Python Client SDK for the Mistral AI API.
-Home-page: https://github.com/mistralai/client-python.git
 Author: Mistral
 Requires-Python: >=3.8
 Classifier: Programming Language :: Python :: 3
@@ -68,6 +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)
@@ -78,6 +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
@@ -118,9 +123,7 @@ with Mistral(
             "content": "Who is the best French painter? Answer in one short sentence.",
             "role": "user",
         },
-    ])
-
-    assert res is not None
+    ], stream=False)
 
     # Handle response
     print(res)
@@ -145,9 +148,7 @@ async def main():
                 "content": "Who is the best French painter? Answer in one short sentence.",
                 "role": "user",
             },
-        ])
-
-        assert res is not None
+        ], stream=False)
 
         # Handle response
         print(res)
@@ -173,8 +174,6 @@ with Mistral(
         "content": open("example.file", "rb"),
     })
 
-    assert res is not None
-
     # Handle response
     print(res)
 ```
@@ -198,8 +197,6 @@ async def main():
             "content": open("example.file", "rb"),
         })
 
-        assert res is not None
-
         # Handle response
         print(res)
 
@@ -224,9 +221,7 @@ with Mistral(
             "content": "Who is the best French painter? Answer in one short sentence.",
             "role": "user",
         },
-    ], agent_id="<value>")
-
-    assert res is not None
+    ], agent_id="<id>", stream=False)
 
     # Handle response
     print(res)
@@ -251,9 +246,7 @@ async def main():
                 "content": "Who is the best French painter? Answer in one short sentence.",
                 "role": "user",
             },
-        ], agent_id="<value>")
-
-        assert res is not None
+        ], agent_id="<id>", stream=False)
 
         # Handle response
         print(res)
@@ -277,9 +270,7 @@ with Mistral(
     res = mistral.embeddings.create(inputs=[
         "Embed this sentence.",
         "As well as this one.",
-    ], model="Wrangler")
-
-    assert res is not None
+    ], model="mistral-embed")
 
     # Handle response
     print(res)
@@ -302,9 +293,7 @@ async def main():
         res = await mistral.embeddings.create_async(inputs=[
             "Embed this sentence.",
             "As well as this one.",
-        ], model="Wrangler")
-
-        assert res is not None
+        ], model="mistral-embed")
 
         # Handle response
         print(res)
@@ -506,9 +495,7 @@ with Mistral(
             "content": "Who is the best French painter? Answer in one short sentence.",
             "role": "user",
         },
-    ])
-
-    assert res is not None
+    ], stream=True)
 
     with res as event_stream:
         for event in event_stream:
@@ -545,8 +532,6 @@ with Mistral(
         "content": open("example.file", "rb"),
     })
 
-    assert res is not None
-
     # Handle response
     print(res)
 
@@ -571,8 +556,6 @@ with Mistral(
     res = mistral.models.list(,
         RetryConfig("backoff", BackoffStrategy(1, 50, 1.1, 100), False))
 
-    assert res is not None
-
     # Handle response
     print(res)
 
@@ -591,8 +574,6 @@ with Mistral(
 
     res = mistral.models.list()
 
-    assert res is not None
-
     # Handle response
     print(res)
 
@@ -634,8 +615,6 @@ with Mistral(
 
         res = mistral.models.list()
 
-        assert res is not None
-
         # Handle response
         print(res)
 
@@ -672,8 +651,6 @@ with Mistral(
 
     res = mistral.models.list()
 
-    assert res is not None
-
     # Handle response
     print(res)
 
@@ -693,8 +670,6 @@ with Mistral(
 
     res = mistral.models.list()
 
-    assert res is not None
-
     # Handle response
     print(res)
 
@@ -804,14 +779,38 @@ with Mistral(
 
     res = mistral.models.list()
 
-    assert res is not None
-
     # Handle response
     print(res)
 
 ```
 <!-- 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-1.3.1 → mistralai-1.5.0}/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
@@ -92,9 +98,7 @@ with Mistral(
             "content": "Who is the best French painter? Answer in one short sentence.",
             "role": "user",
         },
-    ])
-
-    assert res is not None
+    ], stream=False)
 
     # Handle response
     print(res)
@@ -119,9 +123,7 @@ async def main():
                 "content": "Who is the best French painter? Answer in one short sentence.",
                 "role": "user",
             },
-        ])
-
-        assert res is not None
+        ], stream=False)
 
         # Handle response
         print(res)
@@ -147,8 +149,6 @@ with Mistral(
         "content": open("example.file", "rb"),
     })
 
-    assert res is not None
-
     # Handle response
     print(res)
 ```
@@ -172,8 +172,6 @@ async def main():
             "content": open("example.file", "rb"),
         })
 
-        assert res is not None
-
         # Handle response
         print(res)
 
@@ -198,9 +196,7 @@ with Mistral(
             "content": "Who is the best French painter? Answer in one short sentence.",
             "role": "user",
         },
-    ], agent_id="<value>")
-
-    assert res is not None
+    ], agent_id="<id>", stream=False)
 
     # Handle response
     print(res)
@@ -225,9 +221,7 @@ async def main():
                 "content": "Who is the best French painter? Answer in one short sentence.",
                 "role": "user",
             },
-        ], agent_id="<value>")
-
-        assert res is not None
+        ], agent_id="<id>", stream=False)
 
         # Handle response
         print(res)
@@ -251,9 +245,7 @@ with Mistral(
     res = mistral.embeddings.create(inputs=[
         "Embed this sentence.",
         "As well as this one.",
-    ], model="Wrangler")
-
-    assert res is not None
+    ], model="mistral-embed")
 
     # Handle response
     print(res)
@@ -276,9 +268,7 @@ async def main():
         res = await mistral.embeddings.create_async(inputs=[
             "Embed this sentence.",
             "As well as this one.",
-        ], model="Wrangler")
-
-        assert res is not None
+        ], model="mistral-embed")
 
         # Handle response
         print(res)
@@ -480,9 +470,7 @@ with Mistral(
             "content": "Who is the best French painter? Answer in one short sentence.",
             "role": "user",
         },
-    ])
-
-    assert res is not None
+    ], stream=True)
 
     with res as event_stream:
         for event in event_stream:
@@ -519,8 +507,6 @@ with Mistral(
         "content": open("example.file", "rb"),
     })
 
-    assert res is not None
-
     # Handle response
     print(res)
 
@@ -545,8 +531,6 @@ with Mistral(
     res = mistral.models.list(,
         RetryConfig("backoff", BackoffStrategy(1, 50, 1.1, 100), False))
 
-    assert res is not None
-
     # Handle response
     print(res)
 
@@ -565,8 +549,6 @@ with Mistral(
 
     res = mistral.models.list()
 
-    assert res is not None
-
     # Handle response
     print(res)
 
@@ -608,8 +590,6 @@ with Mistral(
 
         res = mistral.models.list()
 
-        assert res is not None
-
         # Handle response
         print(res)
 
@@ -646,8 +626,6 @@ with Mistral(
 
     res = mistral.models.list()
 
-    assert res is not None
-
     # Handle response
     print(res)
 
@@ -667,8 +645,6 @@ with Mistral(
 
     res = mistral.models.list()
 
-    assert res is not None
-
     # Handle response
     print(res)
 
@@ -778,14 +754,38 @@ with Mistral(
 
     res = mistral.models.list()
 
-    assert res is not None
-
     # Handle response
     print(res)
 
 ```
 <!-- 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-1.3.1/packages/mistralai_azure/src/mistralai_azure → mistralai-1.5.0/packages/mistralai_gcp/src/mistralai_gcp}/_hooks/custom_user_agent.py

@@ -19,4 +19,4 @@ class CustomUserAgentHook(BeforeRequestHook):
             PREFIX + current.split(" ")[1]
         )
 
-        return request
+        return request

{mistralai-1.3.1 → mistralai-1.5.0}/packages/mistralai_gcp/src/mistralai_gcp/sdk.py

@@ -30,8 +30,7 @@ def get_model_info(model: str) -> Tuple[str, str]:
     # if the model requiers the legacy fomat, use it, else do nothing.
     if model in LEGACY_MODEL_ID_FORMAT:
         return "-".join(model.split("-")[:-1]), LEGACY_MODEL_ID_FORMAT[model]
-    else:
-        return model, model
+    return model, model
 
 
 

{mistralai-1.3.1 → mistralai-1.5.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "mistralai"
-version = "1.3.1"
+version = "1.5.0"
 description = "Python Client SDK for the Mistral AI API."
 authors = ["Mistral"]
 readme = "README-PYPI.md"

mistralai-1.5.0/src/mistralai/__init__.py

@@ -0,0 +1,18 @@
+"""Code generated by Speakeasy (https://speakeasy.com). DO NOT EDIT."""
+
+from ._version import (
+    __title__,
+    __version__,
+    __openapi_doc_version__,
+    __gen_version__,
+    __user_agent__,
+)
+from .sdk import *
+from .sdkconfiguration import *
+from .models import *
+
+
+VERSION: str = __version__
+OPENAPI_DOC_VERSION = __openapi_doc_version__
+SPEAKEASY_GENERATOR_VERSION = __gen_version__
+USER_AGENT = __user_agent__

{mistralai-1.3.1 → mistralai-1.5.0}/src/mistralai/_version.py

@@ -3,7 +3,10 @@
 import importlib.metadata
 
 __title__: str = "mistralai"
-__version__: str = "1.3.1"
+__version__: str = "1.5.0"
+__openapi_doc_version__: str = "0.0.2"
+__gen_version__: str = "2.497.0"
+__user_agent__: str = "speakeasy-sdk/python 1.5.0 2.497.0 0.0.2 mistralai"
 
 try:
     if __package__ is not None:

{mistralai-1.3.1 → mistralai-1.5.0}/src/mistralai/agents.py

@@ -43,11 +43,14 @@ class Agents(BaseSDK):
         presence_penalty: Optional[float] = None,
         frequency_penalty: Optional[float] = None,
         n: OptionalNullable[int] = UNSET,
+        prediction: Optional[
+            Union[models.Prediction, models.PredictionTypedDict]
+        ] = None,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
         http_headers: Optional[Mapping[str, str]] = None,
-    ) -> Optional[models.ChatCompletionResponse]:
+    ) -> models.ChatCompletionResponse:
         r"""Agents Completion
 
         :param messages: The prompt(s) to generate completions for, encoded as a list of dict with role and content.
@@ -62,6 +65,7 @@ class Agents(BaseSDK):
         :param presence_penalty: presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
         :param frequency_penalty: frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
         :param n: Number of completions to return for each request, input tokens are only billed once.
+        :param prediction:
         :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
@@ -93,6 +97,9 @@ class Agents(BaseSDK):
             presence_penalty=presence_penalty,
             frequency_penalty=frequency_penalty,
             n=n,
+            prediction=utils.get_pydantic_model(
+                prediction, Optional[models.Prediction]
+            ),
             agent_id=agent_id,
         )
 
@@ -138,13 +145,16 @@ class Agents(BaseSDK):
 
         data: Any = None
         if utils.match_response(http_res, "200", "application/json"):
-            return utils.unmarshal_json(
-                http_res.text, Optional[models.ChatCompletionResponse]
-            )
+            return utils.unmarshal_json(http_res.text, models.ChatCompletionResponse)
         if utils.match_response(http_res, "422", "application/json"):
             data = utils.unmarshal_json(http_res.text, models.HTTPValidationErrorData)
             raise models.HTTPValidationError(data=data)
-        if utils.match_response(http_res, ["4XX", "5XX"], "*"):
+        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
+            )
+        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
@@ -191,11 +201,14 @@ class Agents(BaseSDK):
         presence_penalty: Optional[float] = None,
         frequency_penalty: Optional[float] = None,
         n: OptionalNullable[int] = UNSET,
+        prediction: Optional[
+            Union[models.Prediction, models.PredictionTypedDict]
+        ] = None,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
         http_headers: Optional[Mapping[str, str]] = None,
-    ) -> Optional[models.ChatCompletionResponse]:
+    ) -> models.ChatCompletionResponse:
         r"""Agents Completion
 
         :param messages: The prompt(s) to generate completions for, encoded as a list of dict with role and content.
@@ -210,6 +223,7 @@ class Agents(BaseSDK):
         :param presence_penalty: presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
         :param frequency_penalty: frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
         :param n: Number of completions to return for each request, input tokens are only billed once.
+        :param prediction:
         :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
@@ -241,6 +255,9 @@ class Agents(BaseSDK):
             presence_penalty=presence_penalty,
             frequency_penalty=frequency_penalty,
             n=n,
+            prediction=utils.get_pydantic_model(
+                prediction, Optional[models.Prediction]
+            ),
             agent_id=agent_id,
         )
 
@@ -286,13 +303,16 @@ class Agents(BaseSDK):
 
         data: Any = None
         if utils.match_response(http_res, "200", "application/json"):
-            return utils.unmarshal_json(
-                http_res.text, Optional[models.ChatCompletionResponse]
-            )
+            return utils.unmarshal_json(http_res.text, models.ChatCompletionResponse)
         if utils.match_response(http_res, "422", "application/json"):
             data = utils.unmarshal_json(http_res.text, models.HTTPValidationErrorData)
             raise models.HTTPValidationError(data=data)
-        if utils.match_response(http_res, ["4XX", "5XX"], "*"):
+        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
+            )
+        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
@@ -339,11 +359,14 @@ class Agents(BaseSDK):
         presence_penalty: Optional[float] = None,
         frequency_penalty: Optional[float] = None,
         n: OptionalNullable[int] = UNSET,
+        prediction: Optional[
+            Union[models.Prediction, models.PredictionTypedDict]
+        ] = None,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
         http_headers: Optional[Mapping[str, str]] = None,
-    ) -> Optional[eventstreaming.EventStream[models.CompletionEvent]]:
+    ) -> eventstreaming.EventStream[models.CompletionEvent]:
         r"""Stream Agents completion
 
         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.
@@ -360,6 +383,7 @@ class Agents(BaseSDK):
         :param presence_penalty: presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
         :param frequency_penalty: frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
         :param n: Number of completions to return for each request, input tokens are only billed once.
+        :param prediction:
         :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
@@ -391,6 +415,9 @@ class Agents(BaseSDK):
             presence_penalty=presence_penalty,
             frequency_penalty=frequency_penalty,
             n=n,
+            prediction=utils.get_pydantic_model(
+                prediction, Optional[models.Prediction]
+            ),
             agent_id=agent_id,
         )
 
@@ -446,7 +473,12 @@ class Agents(BaseSDK):
             http_res_text = utils.stream_to_text(http_res)
             data = utils.unmarshal_json(http_res_text, models.HTTPValidationErrorData)
             raise models.HTTPValidationError(data=data)
-        if utils.match_response(http_res, ["4XX", "5XX"], "*"):
+        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
+            )
+        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
@@ -493,11 +525,14 @@ class Agents(BaseSDK):
         presence_penalty: Optional[float] = None,
         frequency_penalty: Optional[float] = None,
         n: OptionalNullable[int] = UNSET,
+        prediction: Optional[
+            Union[models.Prediction, models.PredictionTypedDict]
+        ] = None,
         retries: OptionalNullable[utils.RetryConfig] = UNSET,
         server_url: Optional[str] = None,
         timeout_ms: Optional[int] = None,
         http_headers: Optional[Mapping[str, str]] = None,
-    ) -> Optional[eventstreaming.EventStreamAsync[models.CompletionEvent]]:
+    ) -> eventstreaming.EventStreamAsync[models.CompletionEvent]:
         r"""Stream Agents completion
 
         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.
@@ -514,6 +549,7 @@ class Agents(BaseSDK):
         :param presence_penalty: presence_penalty determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative.
         :param frequency_penalty: frequency_penalty penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition.
         :param n: Number of completions to return for each request, input tokens are only billed once.
+        :param prediction:
         :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
@@ -545,6 +581,9 @@ class Agents(BaseSDK):
             presence_penalty=presence_penalty,
             frequency_penalty=frequency_penalty,
             n=n,
+            prediction=utils.get_pydantic_model(
+                prediction, Optional[models.Prediction]
+            ),
             agent_id=agent_id,
         )
 
@@ -600,7 +639,12 @@ class Agents(BaseSDK):
             http_res_text = await utils.stream_to_text_async(http_res)
             data = utils.unmarshal_json(http_res_text, models.HTTPValidationErrorData)
             raise models.HTTPValidationError(data=data)
-        if utils.match_response(http_res, ["4XX", "5XX"], "*"):
+        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
+            )
+        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