From e61883d3d9aa41cac7e695ae159b1ac171cdf42d Mon Sep 17 00:00:00 2001
From: Max <max@Ms-Macbook.local>
Date: Wed, 3 Dec 2025 15:21:36 +0100
Subject: [PATCH 1/2] feat(realtime): add usage field to input audio
 transcription completed event

---
 packages/agents-realtime/src/openaiRealtimeEvents.ts | 12 ++++++++++++
 packages/agents-realtime/src/transportLayerEvents.ts | 10 ++++++++++
 2 files changed, 22 insertions(+)

diff --git a/packages/agents-realtime/src/openaiRealtimeEvents.ts b/packages/agents-realtime/src/openaiRealtimeEvents.ts
index 0228eb31..3ec75a7e 100644
--- a/packages/agents-realtime/src/openaiRealtimeEvents.ts
+++ b/packages/agents-realtime/src/openaiRealtimeEvents.ts
@@ -138,6 +138,18 @@ export const conversationItemInputAudioTranscriptionCompletedEventSchema =
     content_index: z.number(),
     transcript: z.string(),
     logprobs: z.array(z.any()).nullable().optional(),
+    usage: z
+      .object({
+        type: z.literal('tokens'),
+        total_tokens: z.number(),
+        input_tokens: z.number(),
+        input_token_details: z.object({
+          text_tokens: z.number(),
+          audio_tokens: z.number(),
+        }),
+        output_tokens: z.number(),
+      })
+      .optional(),
   });
 
 export const conversationItemInputAudioTranscriptionDeltaEventSchema = z.object(
diff --git a/packages/agents-realtime/src/transportLayerEvents.ts b/packages/agents-realtime/src/transportLayerEvents.ts
index 4a1b260c..258f053b 100644
--- a/packages/agents-realtime/src/transportLayerEvents.ts
+++ b/packages/agents-realtime/src/transportLayerEvents.ts
@@ -44,6 +44,16 @@ export type InputAudioTranscriptionCompletedEvent = {
   type: 'conversation.item.input_audio_transcription.completed';
   item_id: string;
   transcript: string;
+  usage?: {
+    type: 'tokens';
+    total_tokens: number;
+    input_tokens: number;
+    input_token_details: {
+      text_tokens: number;
+      audio_tokens: number;
+    };
+    output_tokens: number;
+  };
 };
 
 export type TransportLayerTranscriptDelta = {

From dfb2976001a468fe1d6592ce5802f12852e1b336 Mon Sep 17 00:00:00 2001
From: Kazuhiro Sera <seratch@openai.com>
Date: Thu, 4 Dec 2025 09:26:09 +0900
Subject: [PATCH 2/2] Create five-cheetahs-work.md

---
 .changeset/five-cheetahs-work.md | 5 +++++
 1 file changed, 5 insertions(+)
 create mode 100644 .changeset/five-cheetahs-work.md

diff --git a/.changeset/five-cheetahs-work.md b/.changeset/five-cheetahs-work.md
new file mode 100644
index 00000000..1c46989e
--- /dev/null
+++ b/.changeset/five-cheetahs-work.md
@@ -0,0 +1,5 @@
+---
+"@openai/agents-realtime": patch
+---
+
+feat(realtime): Add usage field to input audio transcription completed event