fix(api): hoist the session transcript type import out of the split import block

The sessionTranscript type import landed inside the multi-line IProvider
type import instead of after it, splitting the member list mid-block.

Author: jddunn

src/api/generateText.ts

@@ -80,12 +80,12 @@ class LLMProviderCircuitOpenError extends Error {
   }
 }
 import type { IModelRouter, ModelRouteParams } from '../core/llm/routing/IModelRouter.js';
-import type { SessionTranscriptMessage } from './sessionTranscript.js';
 import type {
   MessageContent,
   MessageContentPart,
   CacheDiagnostics,
 } from '../core/llm/providers/IProvider.js';
+import type { SessionTranscriptMessage } from './sessionTranscript.js';
 
 // Re-export multimodal types for downstream consumers
 export type { MessageContent, MessageContentPart };
@@ -1829,28 +1829,6 @@ export async function generateText(opts: GenerateTextOptions): Promise<GenerateT
               totalTokens: stepResponse.usage?.totalTokens,
               costUSD: stepResponse.usage?.costUSD,
             });
-            // Per-step GenAI semconv attrs (queue item: step-span wiring).
-            // ModelUsage's cache fields are the *InputTokens spellings —
-            // mapped here onto the ApiUsageLike names the helper reads.
-            attachGenAiAttributes(stepSpan, {
-              providerName: resolved.providerId,
-              operationName: 'chat',
-              requestModel: resolved.modelId,
-              ...(typeof stepResponse.modelId === 'string' && stepResponse.modelId
-                ? { responseModel: stepResponse.modelId }
-                : {}),
-              ...(opts.serviceTier !== undefined ? { requestServiceTier: opts.serviceTier } : {}),
-              ...(typeof stepResponse.serviceTier === 'string'
-                ? { responseServiceTier: stepResponse.serviceTier }
-                : {}),
-              usage: {
-                promptTokens: stepResponse.usage?.promptTokens,
-                completionTokens: stepResponse.usage?.completionTokens,
-                inclusiveInputTokens: stepResponse.usage?.inclusiveInputTokens,
-                cacheReadTokens: stepResponse.usage?.cacheReadInputTokens,
-                cacheCreationTokens: stepResponse.usage?.cacheCreationInputTokens,
-              },
-            });
             return stepResponse;
           }
         );