feat(client): forward max_retries through Client/ClientV2 constructors (#779)

* feat(client): forward max_retries through Client/ClientV2 constructors

The base BaseCohere/AsyncBaseCohere constructors accept `max_retries`,
but the hand-maintained Client, AsyncClient, ClientV2 and AsyncClientV2
override constructors never exposed it, so it could not be set at the
client level (only per-request via request_options).

Thread `max_retries: Optional[int] = None` through all four override
constructors, forwarding to the respective super().__init__. The None
default preserves the existing behavior (base applies its default of 2).

* chore: bump version 7.0.4 -> 7.0.5

---------

Co-authored-by: jsklan <jsklan.development@gmail.com>

Author: fern-support

pyproject.toml

@@ -4,7 +4,7 @@ dynamic = ["version"]
 
 [tool.poetry]
 name = "cohere"
-version = "7.0.4"
+version = "7.0.5"
 description = ""
 readme = "README.md"
 authors = []

src/cohere/client.py

@@ -139,6 +139,7 @@ def __init__(
         environment: ClientEnvironment = ClientEnvironment.PRODUCTION,
         client_name: typing.Optional[str] = None,
         timeout: typing.Optional[float] = None,
+        max_retries: typing.Optional[int] = None,
         httpx_client: typing.Optional[httpx.Client] = None,
         thread_pool_executor: ThreadPoolExecutor = ThreadPoolExecutor(64),
         log_warning_experimental_features: bool = True,
@@ -157,6 +158,7 @@ def __init__(
             client_name=client_name,
             token=api_key,
             timeout=timeout,
+            max_retries=max_retries,
             httpx_client=httpx_client,
         )
 
@@ -386,6 +388,7 @@ def __init__(
         environment: ClientEnvironment = ClientEnvironment.PRODUCTION,
         client_name: typing.Optional[str] = None,
         timeout: typing.Optional[float] = None,
+        max_retries: typing.Optional[int] = None,
         httpx_client: typing.Optional[httpx.AsyncClient] = None,
         thread_pool_executor: ThreadPoolExecutor = ThreadPoolExecutor(64),
         log_warning_experimental_features: bool = True,
@@ -404,6 +407,7 @@ def __init__(
             client_name=client_name,
             token=api_key,
             timeout=timeout,
+            max_retries=max_retries,
             httpx_client=httpx_client,
         )
 

src/cohere/client_v2.py

@@ -38,6 +38,7 @@ def __init__(
         environment: ClientEnvironment = ClientEnvironment.PRODUCTION,
         client_name: typing.Optional[str] = None,
         timeout: typing.Optional[float] = None,
+        max_retries: typing.Optional[int] = None,
         httpx_client: typing.Optional[httpx.Client] = None,
         thread_pool_executor: ThreadPoolExecutor = ThreadPoolExecutor(64),
         log_warning_experimental_features: bool = True,
@@ -49,6 +50,7 @@ def __init__(
             environment=environment,
             client_name=client_name,
             timeout=timeout,
+            max_retries=max_retries,
             httpx_client=httpx_client,
             thread_pool_executor=thread_pool_executor,
             log_warning_experimental_features=log_warning_experimental_features,
@@ -71,6 +73,7 @@ def __init__(
         environment: ClientEnvironment = ClientEnvironment.PRODUCTION,
         client_name: typing.Optional[str] = None,
         timeout: typing.Optional[float] = None,
+        max_retries: typing.Optional[int] = None,
         httpx_client: typing.Optional[httpx.AsyncClient] = None,
         thread_pool_executor: ThreadPoolExecutor = ThreadPoolExecutor(64),
         log_warning_experimental_features: bool = True,
@@ -82,6 +85,7 @@ def __init__(
             environment=environment,
             client_name=client_name,
             timeout=timeout,
+            max_retries=max_retries,
             httpx_client=httpx_client,
             thread_pool_executor=thread_pool_executor,
             log_warning_experimental_features=log_warning_experimental_features,

src/cohere/core/client_wrapper.py

@@ -31,12 +31,12 @@ def get_headers(self) -> typing.Dict[str, str]:
         import platform
 
         headers: typing.Dict[str, str] = {
-            "User-Agent": "cohere/7.0.4",
+            "User-Agent": "cohere/7.0.5",
             "X-Fern-Language": "Python",
             "X-Fern-Runtime": f"python/{platform.python_version()}",
             "X-Fern-Platform": f"{platform.system().lower()}/{platform.release()}",
             "X-Fern-SDK-Name": "cohere",
-            "X-Fern-SDK-Version": "7.0.4",
+            "X-Fern-SDK-Version": "7.0.5",
             **(self.get_custom_headers() or {}),
         }
         if self._client_name is not None: