codex: use item done for realtime transcript done

codex-rs/codex-api/src/endpoint/realtime_websocket/methods.rs

@@ -1136,6 +1136,35 @@ mod tests {
         );
     }
 
+    #[test]
+    fn parse_realtime_v2_item_done_output_text_event() {
+        let payload = json!({
+            "type": "conversation.item.done",
+            "item": {
+                "id": "item_output_1",
+                "type": "message",
+                "role": "assistant",
+                "content": [
+                    {"type": "output_text", "text": "hello"},
+                    {"type": "output_text", "text": " world"}
+                ]
+            }
+        })
+        .to_string();
+
+        assert_eq!(
+            parse_realtime_event(payload.as_str(), RealtimeEventParser::RealtimeV2),
+            Some(RealtimeEvent::OutputTranscriptDone(
+                RealtimeTranscriptDone {
+                    text: "hello world".to_string(),
+                    item_id: Some("item_output_1".to_string()),
+                    output_index: None,
+                    content_index: None,
+                }
+            ))
+        );
+    }
+
     #[test]
     fn complete_transcript_entry_replaces_current_part_only() {
         let mut state = ActiveTranscriptState::default();

codex-rs/codex-api/src/endpoint/realtime_websocket/protocol_v2.rs

@@ -10,6 +10,7 @@ use codex_protocol::protocol::RealtimeInputAudioSpeechStarted;
 use codex_protocol::protocol::RealtimeResponseCancelled;
 use codex_protocol::protocol::RealtimeResponseCreated;
 use codex_protocol::protocol::RealtimeResponseDone;
+use codex_protocol::protocol::RealtimeTranscriptDone;
 use serde_json::Map as JsonMap;
 use serde_json::Value;
 use tracing::debug;
@@ -128,12 +129,48 @@ fn parse_conversation_item_done_event(parsed: &Value) -> Option<RealtimeEvent> {
         return Some(handoff);
     }
 
+    if let Some(transcript_done) = parse_item_done_transcript(item) {
+        return Some(transcript_done);
+    }
+
     item.get("id")
         .and_then(Value::as_str)
         .map(str::to_string)
         .map(|item_id| RealtimeEvent::ConversationItemDone { item_id })
 }
 
+fn parse_item_done_transcript(item: &JsonMap<String, Value>) -> Option<RealtimeEvent> {
+    let role = item.get("role").and_then(Value::as_str)?;
+    let text = item
+        .get("content")
+        .and_then(Value::as_array)?
+        .iter()
+        .filter_map(item_content_text)
+        .collect::<String>();
+    if text.is_empty() {
+        return None;
+    }
+
+    let done = RealtimeTranscriptDone {
+        text,
+        item_id: item.get("id").and_then(Value::as_str).map(str::to_string),
+        output_index: None,
+        content_index: None,
+    };
+    match role {
+        "user" => Some(RealtimeEvent::InputTranscriptDone(done)),
+        "assistant" => Some(RealtimeEvent::OutputTranscriptDone(done)),
+        _ => None,
+    }
+}
+
+fn item_content_text(content: &Value) -> Option<&str> {
+    content
+        .get("text")
+        .or_else(|| content.get("transcript"))
+        .and_then(Value::as_str)
+}
+
 fn parse_handoff_requested_event(item: &JsonMap<String, Value>) -> Option<RealtimeEvent> {
     let item_type = item.get("type").and_then(Value::as_str);
     let item_name = item.get("name").and_then(Value::as_str);