Audio cost calculation

Ironclad · Feb 2, 2025 · 0de1731 · 0de1731
1 parent e6fd847
commit 0de1731
Show file tree

Hide file tree

Showing 2 changed files with 82 additions and 0 deletions.
diff --git a/packages/core/src/model/nodes/ChatNode.ts b/packages/core/src/model/nodes/ChatNode.ts
@@ -1160,6 +1160,55 @@ export class ChatNodeImpl extends NodeImpl<ChatNode> {
 
             output['duration' as PortId] = { type: 'number', value: Date.now() - startTime };
 
+            if (response.usage) {
+              output['usage' as PortId] = {
+                type: 'object',
+                value: response.usage,
+              };
+
+              const costs =
+                finalModel in openaiModels ? openaiModels[finalModel as keyof typeof openaiModels].cost : undefined;
+
+              const promptCostPerThousand = costs?.prompt ?? 0;
+              const completionCostPerThousand = costs?.completion ?? 0;
+              const audioPromptCostPerThousand = costs
+                ? 'audioPrompt' in costs
+                  ? (costs.audioPrompt as number)
+                  : 0
+                : 0;
+              const audioCompletionCostPerThousand = costs
+                ? 'audioCompletion' in costs
+                  ? (costs.audioCompletion as number)
+                  : 0
+                : 0;
+
+              const promptCost = getCostForTokens(
+                response.usage.prompt_tokens_details.text_tokens,
+                'prompt',
+                promptCostPerThousand,
+              );
+              const completionCost = getCostForTokens(
+                response.usage.completion_tokens_details.text_tokens,
+                'completion',
+                completionCostPerThousand,
+              );
+              const audioPromptCost = getCostForTokens(
+                response.usage.prompt_tokens_details.audio_tokens,
+                'prompt',
+                audioPromptCostPerThousand,
+              );
+              const audioCompletionCost = getCostForTokens(
+                response.usage.completion_tokens_details.audio_tokens,
+                'completion',
+                audioCompletionCostPerThousand,
+              );
+
+              output['cost' as PortId] = {
+                type: 'number',
+                value: promptCost + completionCost + audioPromptCost + audioCompletionCost,
+              };
+            }
+
             Object.freeze(output);
             cache.set(cacheKey, output);
 

diff --git a/packages/core/src/utils/openai.ts b/packages/core/src/utils/openai.ts
@@ -9,6 +9,9 @@ export type OpenAIModel = {
   cost: {
     prompt: number;
     completion: number;
+
+    audioPrompt?: number;
+    audioCompletion?: number;
   };
   displayName: string;
 };
@@ -171,6 +174,8 @@ export const openaiModels = {
     cost: {
       prompt: 0.0025,
       completion: 0.01,
+      audioPrompt: 0.04,
+      audioCompletion: 0.08,
     },
     displayName: 'GPT-4o Audio (Preview)',
   },
@@ -365,6 +370,34 @@ export type ChatCompletionResponse = {
 
     /** Total number of tokens used in the request (prompt + completion). */
     total_tokens: number;
+
+    prompt_tokens_details: {
+      /** Number of tokens used from the cache. */
+      cached_tokens: number;
+
+      /** Number of tokens used for audio. */
+      audio_tokens: number;
+
+      text_tokens: number;
+
+      image_tokens: number;
+    };
+
+    completion_tokens_details: {
+      /** Number of tokens used for reasoning. */
+      reasoning_tokens: number;
+
+      /** Number of tokens used for audio. */
+      audio_tokens: number;
+
+      /** Number of tokens used for accepted predictions. */
+      accepted_prediction_tokens: number;
+
+      /** Number of tokens used for rejected predictions. */
+      rejected_prediction_tokens: number;
+
+      text_tokens: number;
+    };
   };
 
   /** A list of chat completion choices. Can be more than one if n is greater than 1. */