feat(providers): allow xhigh/max reasoning_effort on the OpenAI-compatible path

The per-agent options builder only emitted WithReasoning for low/medium/high,
and the langchaingo fork clamped xhigh/max to high in GetEffort (treating them
as adaptive-only). OpenAI GPT-5.5 and GLM-5.2 now expose xhigh/max as real
reasoning_effort levels, so that clamp encodes a stale invariant.

Extend the effort switch to pass xhigh/max through; the fork's GetEffort clamp
is removed in the local langchaingo (propagated via vendor) so the OpenAI
transport emits the real level. Adaptive (anthropic/bedrock) and budget paths
are unaffected — GetEffort is called only by the OpenAI transport. Validity is
gated per model by ModelReasoningInfo.Efforts.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/pkg/providers/pconfig/config.go

@@ -724,7 +724,7 @@ func (ac *AgentConfig) BuildOptions() []llms.CallOption {
 			}
 		default:
 			switch ac.Reasoning.Effort {
-			case llms.ReasoningLow, llms.ReasoningMedium, llms.ReasoningHigh:
+			case llms.ReasoningLow, llms.ReasoningMedium, llms.ReasoningHigh, llms.ReasoningXHigh, llms.ReasoningMax:
 				options = append(options, llms.WithReasoning(ac.Reasoning.Effort, 0))
 			}
 		}

backend/pkg/providers/pconfig/reasoning_effort_test.go

@@ -0,0 +1,37 @@
+package pconfig
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"github.com/vxcontrol/langchaingo/llms"
+	"gopkg.in/yaml.v3"
+)
+
+func TestReasoningConfig_GetEffortPreservesXHighAndMax(t *testing.T) {
+	for _, effort := range []llms.ReasoningEffort{llms.ReasoningXHigh, llms.ReasoningMax} {
+		rc := llms.ReasoningConfig{Effort: effort}
+		if got := rc.GetEffort(8192); got != effort {
+			t.Errorf("GetEffort(effort=%q) = %q, want %q: the xhigh/max clamp must stay removed so OpenAI-compatible reasoning_effort carries the real level", effort, got, effort)
+		}
+	}
+}
+
+func TestProviderConfig_EmitsTopReasoningEffortOnEffortPath(t *testing.T) {
+	for _, effort := range []string{"low", "medium", "high", "xhigh", "max"} {
+		t.Run(effort, func(t *testing.T) {
+			var ac AgentConfig
+			require.NoError(t, yaml.Unmarshal([]byte("model: gpt-5.5\nreasoning:\n  effort: "+effort+"\n"), &ac))
+
+			pc := &ProviderConfig{Simple: &ac}
+			var applied llms.CallOptions
+			for _, opt := range pc.GetOptionsForType(OptionsTypeSimple) {
+				opt(&applied)
+			}
+
+			require.NotNil(t, applied.Reasoning, "effort %q must emit a reasoning option on the non-adaptive path", effort)
+			assert.Equal(t, llms.ReasoningEffort(effort), applied.Reasoning.Effort, "the configured effort must reach the call options unchanged")
+		})
+	}
+}