refactor: single computed EffectiveReasoning() on LLMResponse, prefer Reasoning over ReasoningContent (#1)

* Add EffectiveReasoning() to LLMResponse and use it in loop.go

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: CzBiX <760097+CzBiX@users.noreply.github.com>

pkg/agent/loop.go

@@ -1091,9 +1091,10 @@ func (al *AgentLoop) runLLMIteration(
 			return "", iteration, fmt.Errorf("LLM call failed after retries: %w", err)
 		}
 
+		effectiveReasoning := response.EffectiveReasoning()
 		go al.handleReasoning(
 			ctx,
-			response.Reasoning,
+			effectiveReasoning,
 			opts.Channel,
 			al.targetReasoningChannelID(opts.Channel),
 		)
@@ -1104,7 +1105,7 @@ func (al *AgentLoop) runLLMIteration(
 				"iteration":      iteration,
 				"content_chars":  len(response.Content),
 				"tool_calls":     len(response.ToolCalls),
-				"reasoning":      response.Reasoning,
+				"reasoning":      effectiveReasoning,
 				"target_channel": al.targetReasoningChannelID(opts.Channel),
 				"channel":        opts.Channel,
 			})

pkg/providers/protocoltypes/types.go

@@ -34,6 +34,14 @@ type LLMResponse struct {
 	ReasoningDetails []ReasoningDetail `json:"reasoning_details"`
 }
 
+// EffectiveReasoning returns Reasoning if non-empty, otherwise ReasoningContent.
+func (r *LLMResponse) EffectiveReasoning() string {
+	if r.Reasoning != "" {
+		return r.Reasoning
+	}
+	return r.ReasoningContent
+}
+
 type ReasoningDetail struct {
 	Format string `json:"format"`
 	Index  int    `json:"index"`