Skip to content

CubeLoop — a Pythonic, async-native agent harness core designed for high performance, readability, and production-grade persistence.

fix(providers): stop emitting stale "minimal" reasoning_effort for OpenAI

xf gongc63c545b1612
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 3f9263c..ecf06da 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -7,6 +7,15 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
## [Unreleased]
+### Fixed
+
+- **OpenAI `chat_completions`/`responses` capability profiles no longer emit
+ `reasoning_effort: "minimal"`.** OpenAI renamed the lowest reasoning-effort
+ tier to `"none"` (gpt-5.5+), and some OpenAI-compatible proxies reject the
+ now-stale `"minimal"` value outright. The built-in "off" mode payload and
+ the `minimal` → wire-value mapping in `_OPENAI_EFFORT_VALUES` both now emit
+ `"none"`.
+
## [0.13.2] - 2026-07-23
### Fixed
diff --git a/cubepi/providers/reasoning_profiles.py b/cubepi/providers/reasoning_profiles.py
index b797516..287ccfd 100644
--- a/cubepi/providers/reasoning_profiles.py
+++ b/cubepi/providers/reasoning_profiles.py
@@ -7,7 +7,10 @@ from cubepi.providers.capability import CapabilityDescriptor, ReasoningCapabilit
_OPENAI_EFFORT_VALUES: dict[ReasoningEffort, str] = {
- "minimal": "minimal",
+ # OpenAI renamed the lowest reasoning_effort tier from "minimal" to
+ # "none" (gpt-5.5+); some OpenAI-compatible backends reject "minimal"
+ # outright.
+ "minimal": "none",
"low": "low",
"medium": "medium",
"high": "high",
@@ -50,7 +53,7 @@ def get_capability_profile(
_PROFILES: dict[tuple[str, str], CapabilityDescriptor] = {
("openai", "chat_completions"): CapabilityDescriptor(
reasoning=ReasoningCapability(
- mode_payloads={"off": {"reasoning_effort": "minimal"}},
+ mode_payloads={"off": {"reasoning_effort": "none"}},
effort_path="reasoning_effort",
effort_values=_OPENAI_EFFORT_VALUES,
apply_effort_when_off=False,
@@ -59,7 +62,7 @@ _PROFILES: dict[tuple[str, str], CapabilityDescriptor] = {
),
("openai", "responses"): CapabilityDescriptor(
reasoning=ReasoningCapability(
- mode_payloads={"off": {"reasoning": {"effort": "minimal"}}},
+ mode_payloads={"off": {"reasoning": {"effort": "none"}}},
effort_path="reasoning.effort",
effort_values=_OPENAI_EFFORT_VALUES,
summary_path="reasoning.summary",
diff --git a/tests/providers/test_openai_capability.py b/tests/providers/test_openai_capability.py
index 751feea..d2919d8 100644
--- a/tests/providers/test_openai_capability.py
+++ b/tests/providers/test_openai_capability.py
@@ -161,14 +161,14 @@ async def test_openai_chat_default_profile_writes_reasoning_effort() -> None:
@pytest.mark.asyncio
-async def test_openai_chat_off_writes_minimal_effort() -> None:
+async def test_openai_chat_off_writes_none_effort() -> None:
payload = await _capture_payload_openai(
OpenAIProvider(api_key="x", base_url="http://example"),
_model(reasoning=True),
reasoning=ReasoningControl(mode="off", effort="minimal", summary="none"),
)
- assert payload["reasoning_effort"] == "minimal"
+ assert payload["reasoning_effort"] == "none"
@pytest.mark.asyncio
diff --git a/tests/providers/test_openai_responses.py b/tests/providers/test_openai_responses.py
index b1fbaa4..e70bde0 100644
--- a/tests/providers/test_openai_responses.py
+++ b/tests/providers/test_openai_responses.py
@@ -854,7 +854,7 @@ class TestReasoningParams:
@pytest.mark.asyncio
async def test_no_reasoning_when_off(self):
- """mode=off maps to minimal effort for reasoning-only Responses models."""
+ """mode=off maps to "none" effort for reasoning-only Responses models."""
events = [
_make_event(
"response.completed",
@@ -898,7 +898,7 @@ class TestReasoningParams:
await ms.result()
call_kwargs = mock_client.responses.create.call_args[1]
- assert call_kwargs["reasoning"] == {"effort": "minimal"}
+ assert call_kwargs["reasoning"] == {"effort": "none"}
assert "include" not in call_kwargs
@pytest.mark.asyncio
diff --git a/tests/providers/test_openai_responses_capability.py b/tests/providers/test_openai_responses_capability.py
index 68ccb42..8788b5d 100644
--- a/tests/providers/test_openai_responses_capability.py
+++ b/tests/providers/test_openai_responses_capability.py
@@ -153,14 +153,14 @@ async def test_default_profile_writes_reasoning_effort():
@pytest.mark.asyncio
-async def test_default_profile_off_writes_minimal_reasoning():
+async def test_default_profile_off_writes_none_reasoning():
p = OpenAIResponsesProvider(api_key="x")
payload = await _capture_payload_responses(
p,
_model(reasoning=True),
reasoning=ReasoningControl(mode="off", effort="minimal", summary="none"),
)
- assert payload["reasoning"] == {"effort": "minimal"}
+ assert payload["reasoning"] == {"effort": "none"}
# ---------------------------------------------------------------------------
diff --git a/tests/providers/test_reasoning_capability.py b/tests/providers/test_reasoning_capability.py
index 1b6e4cd..373a940 100644
--- a/tests/providers/test_reasoning_capability.py
+++ b/tests/providers/test_reasoning_capability.py
@@ -31,7 +31,7 @@ def test_apply_reasoning_writes_effort_when_off_for_chat_profile():
ReasoningControl(mode="off", effort="minimal"),
)
- assert payload == {"reasoning_effort": "minimal"}
+ assert payload == {"reasoning_effort": "none"}
def test_apply_reasoning_writes_nested_summary_and_include_payload():
diff --git a/uv.lock b/uv.lock
index 6acfa02..4ad5161 100644
--- a/uv.lock
+++ b/uv.lock
@@ -522,7 +522,7 @@ wheels = [
[[package]]
name = "cubepi"
-version = "0.13.1"
+version = "0.13.2"
source = { editable = "." }
dependencies = [
{ name = "anthropic" },