Honor SDK retry configuration

src/stirrup/clients/chat_completions_client.py

@@ -12,14 +12,7 @@
 from time import perf_counter
 from typing import Any
 
-from openai import (
-    APIConnectionError,
-    APITimeoutError,
-    AsyncOpenAI,
-    InternalServerError,
-    RateLimitError,
-)
-from tenacity import retry, retry_if_exception_type, stop_after_attempt, wait_exponential
+from openai import AsyncOpenAI
 
 from stirrup.clients.utils import to_openai_messages, to_openai_tools
 from stirrup.core.exceptions import ContextOverflowError
@@ -47,7 +40,7 @@ class ChatCompletionsClient(LLMClient):
     Supports custom base_url for OpenAI-compatible providers (vLLM, Ollama,
     Azure OpenAI, local models, etc.).
 
-    Includes automatic retries for transient failures and token usage tracking.
+    Delegates retries for transient failures to the OpenAI SDK and tracks token usage.
 
     Example:
         >>> # Standard OpenAI usage
@@ -114,28 +107,13 @@ def model_slug(self) -> str:
         """Model identifier."""
         return self._model
 
-    @retry(
-        retry=retry_if_exception_type(
-            (
-                APIConnectionError,
-                APITimeoutError,
-                RateLimitError,
-                InternalServerError,
-            )
-        ),
-        stop=stop_after_attempt(3),
-        wait=wait_exponential(multiplier=1, min=1, max=10),
-    )
     async def generate(
         self,
         messages: list[ChatMessage],
         tools: dict[str, Tool],
     ) -> AssistantMessage:
         """Generate assistant response with optional tool calls.
 
-        Retries up to 3 times on transient errors (connection, timeout, rate limit,
-        internal server errors) with exponential backoff.
-
         Args:
             messages: List of conversation messages.
             tools: Dictionary mapping tool names to Tool objects.

src/stirrup/clients/open_responses_client.py

@@ -10,14 +10,7 @@
 from time import perf_counter
 from typing import Any
 
-from openai import (
-    APIConnectionError,
-    APITimeoutError,
-    AsyncOpenAI,
-    InternalServerError,
-    RateLimitError,
-)
-from tenacity import retry, retry_if_exception_type, stop_after_attempt, wait_exponential
+from openai import AsyncOpenAI
 
 from stirrup.core.exceptions import ContextOverflowError
 from stirrup.core.models import (
@@ -263,7 +256,7 @@ class OpenResponsesClient(LLMClient):
     Supports custom base_url for OpenAI-compatible providers that implement
     the Responses API.
 
-    Includes automatic retries for transient failures and token usage tracking.
+    Delegates retries for transient failures to the OpenAI SDK and tracks token usage.
 
     Example:
         >>> # Standard OpenAI usage
@@ -303,6 +296,7 @@ def __init__(
                 (e.g., 'low', 'medium', 'high'). Only used with o1/o3 style models.
             timeout: Request timeout in seconds. If None, uses OpenAI SDK default.
             max_retries: Number of retries for transient errors. Defaults to 2.
+                The OpenAI SDK handles retries internally.
             instructions: Default system-level instructions. Can be overridden by
                 SystemMessage in the messages list.
             kwargs: Additional arguments passed to responses.create().
@@ -338,28 +332,13 @@ def model_slug(self) -> str:
         """Model identifier."""
         return self._model
 
-    @retry(
-        retry=retry_if_exception_type(
-            (
-                APIConnectionError,
-                APITimeoutError,
-                RateLimitError,
-                InternalServerError,
-            )
-        ),
-        stop=stop_after_attempt(3),
-        wait=wait_exponential(multiplier=1, min=1, max=10),
-    )
     async def generate(
         self,
         messages: list[ChatMessage],
         tools: dict[str, Tool],
     ) -> AssistantMessage:
         """Generate assistant response with optional tool calls using Responses API.
 
-        Retries up to 3 times on transient errors (connection, timeout, rate limit,
-        internal server errors) with exponential backoff.
-
         Args:
             messages: List of conversation messages.
             tools: Dictionary mapping tool names to Tool objects.

tests/test_openai_sdk_retries.py

@@ -0,0 +1,47 @@
+"""Behavioral tests for OpenAI SDK retry configuration."""
+
+import httpx
+import pytest
+from openai import InternalServerError
+
+from stirrup.clients.chat_completions_client import ChatCompletionsClient
+from stirrup.clients.open_responses_client import OpenResponsesClient
+from stirrup.core.models import UserMessage
+
+
+@pytest.mark.parametrize(
+    "client_class",
+    [
+        pytest.param(ChatCompletionsClient, id="chat_completions"),
+        pytest.param(OpenResponsesClient, id="open_responses"),
+    ],
+)
+async def test_max_retries_controls_sdk_request_count(
+    client_class: type[ChatCompletionsClient] | type[OpenResponsesClient],
+) -> None:
+    request_count = 0
+
+    def fail(request: httpx.Request) -> httpx.Response:
+        nonlocal request_count
+        request_count += 1
+        return httpx.Response(
+            500,
+            headers={"retry-after-ms": "1"},
+            json={"error": {"message": "retry", "type": "server_error"}},
+            request=request,
+        )
+
+    client = client_class(model="gpt-4o", api_key="test-key", max_retries=1)
+    original_sdk_client = client._client  # noqa: SLF001
+    client._client = original_sdk_client.with_options(  # noqa: SLF001
+        http_client=httpx.AsyncClient(transport=httpx.MockTransport(fail))
+    )
+    await original_sdk_client.close()
+
+    try:
+        with pytest.raises(InternalServerError):
+            await client.generate([UserMessage(content="hello")], {})
+    finally:
+        await client._client.close()  # noqa: SLF001
+
+    assert request_count == 2