Persist timestamped realtime frames before acknowledging completion (#61473)

## What changed

- Record voice transcripts as `frame` items with a `voice_transcript` channel for eligible threads, while retaining legacy transcript support and the existing app-server live and mixed timeline representation.
- Assign first-observed timestamps that never move backward, retain them through completion, and restore the clock floor from history and compaction checkpoints.
- Flush realtime records before completion notifications, including file and initial directory durability barriers for local rollouts.
- Generate a distinct default session ID for each new call, use the call ID when attaching to an existing call, and derive opening record IDs from session IDs.
- Add protocol types for turn-input acceptance and session-end input receipts, excluding those receipts from live notifications and mixed timelines. Update generated schemas and SDK types.

## Testing

Add coverage for frame serialization and required voice payloads, clock rollback and checkpoint restoration, durability barrier failures and retries, per-call session IDs, and persisted transcript availability immediately after completion.

GitOrigin-RevId: 3f6632b517d563861aee085b723475b8945c391c

sdk/python/src/openai_codex/generated/v2_all.py

@@ -6423,6 +6423,11 @@ class ThreadRealtimeErrorNotification(BaseModel):
     thread_id: Annotated[str, Field(alias="threadId")]
 
 
+class ThreadRealtimeHandoffMode(Enum):
+    start = "start"
+    steer = "steer"
+
+
 class ThreadRealtimeInitialItem(BaseModel):
     model_config = ConfigDict(
         populate_by_name=True,
@@ -6436,6 +6441,13 @@ class RealtimeSessionStartedThreadRealtimeItem(BaseModel):
         populate_by_name=True,
     )
     id: str
+    observed_at_ms: Annotated[
+        int | None,
+        Field(
+            alias="observedAtMs",
+            description="Canonical observation time in Unix milliseconds for eligible voice calls.",
+        ),
+    ] = None
     realtime_session_id: Annotated[str, Field(alias="realtimeSessionId")]
     type: Annotated[
         Literal["realtimeSessionStarted"],
@@ -6448,6 +6460,13 @@ class BemItemPromotedThreadRealtimeItem(BaseModel):
         populate_by_name=True,
     )
     id: str
+    observed_at_ms: Annotated[
+        int | None,
+        Field(
+            alias="observedAtMs",
+            description="Canonical observation time in Unix milliseconds for eligible voice calls.",
+        ),
+    ] = None
     realtime_session_id: Annotated[str, Field(alias="realtimeSessionId")]
     item_id: str
     presentation: ThreadRealtimeBemItemPresentation
@@ -6457,6 +6476,51 @@ class BemItemPromotedThreadRealtimeItem(BaseModel):
     ]
 
 
+class TurnInputAcceptedThreadRealtimeItem(BaseModel):
+    model_config = ConfigDict(
+        populate_by_name=True,
+    )
+    id: str
+    observed_at_ms: Annotated[
+        int | None,
+        Field(
+            alias="observedAtMs",
+            description="Canonical observation time in Unix milliseconds for eligible voice calls.",
+        ),
+    ] = None
+    realtime_session_id: Annotated[str, Field(alias="realtimeSessionId")]
+    handoff_id: str
+    input: str
+    input_item_id: str
+    mode: ThreadRealtimeHandoffMode
+    source_event_ids: list[str]
+    turn_id: str
+    type: Annotated[
+        Literal["turnInputAccepted"], Field(title="TurnInputAcceptedThreadRealtimeItemType")
+    ]
+
+
+class SessionEndInputSavedThreadRealtimeItem(BaseModel):
+    model_config = ConfigDict(
+        populate_by_name=True,
+    )
+    id: str
+    observed_at_ms: Annotated[
+        int | None,
+        Field(
+            alias="observedAtMs",
+            description="Canonical observation time in Unix milliseconds for eligible voice calls.",
+        ),
+    ] = None
+    realtime_session_id: Annotated[str, Field(alias="realtimeSessionId")]
+    input: str
+    input_item_id: str
+    source_event_ids: list[str]
+    type: Annotated[
+        Literal["sessionEndInputSaved"], Field(title="SessionEndInputSavedThreadRealtimeItemType")
+    ]
+
+
 class ThreadRealtimeItemAddedNotification(BaseModel):
     model_config = ConfigDict(
         populate_by_name=True,
@@ -6578,6 +6642,14 @@ class ThreadRealtimeTranscriptRole(Enum):
     assistant = "assistant"
 
 
+class ThreadRealtimeVoiceTranscript(BaseModel):
+    model_config = ConfigDict(
+        populate_by_name=True,
+    )
+    role: ThreadRealtimeTranscriptRole
+    text: str
+
+
 class ThreadResumeParams(BaseModel):
     model_config = ConfigDict(
         populate_by_name=True,
@@ -11238,11 +11310,42 @@ class ThreadReadStateChangedNotification(BaseModel):
     thread_id: Annotated[str, Field(alias="threadId")]
 
 
+class ThreadRealtimeFrameChannels(BaseModel):
+    model_config = ConfigDict(
+        populate_by_name=True,
+    )
+    voice_transcript: ThreadRealtimeVoiceTranscript
+
+
+class FrameThreadRealtimeItem(BaseModel):
+    model_config = ConfigDict(
+        populate_by_name=True,
+    )
+    id: str
+    observed_at_ms: Annotated[
+        int | None,
+        Field(
+            alias="observedAtMs",
+            description="Canonical observation time in Unix milliseconds for eligible voice calls.",
+        ),
+    ] = None
+    realtime_session_id: Annotated[str, Field(alias="realtimeSessionId")]
+    channels: ThreadRealtimeFrameChannels
+    type: Annotated[Literal["frame"], Field(title="FrameThreadRealtimeItemType")]
+
+
 class TranscriptSegmentThreadRealtimeItem(BaseModel):
     model_config = ConfigDict(
         populate_by_name=True,
     )
     id: str
+    observed_at_ms: Annotated[
+        int | None,
+        Field(
+            alias="observedAtMs",
+            description="Canonical observation time in Unix milliseconds for eligible voice calls.",
+        ),
+    ] = None
     realtime_session_id: Annotated[str, Field(alias="realtimeSessionId")]
     role: ThreadRealtimeTranscriptRole
     text: str
@@ -11256,6 +11359,13 @@ class RealtimeSessionClosedThreadRealtimeItem(BaseModel):
         populate_by_name=True,
     )
     id: str
+    observed_at_ms: Annotated[
+        int | None,
+        Field(
+            alias="observedAtMs",
+            description="Canonical observation time in Unix milliseconds for eligible voice calls.",
+        ),
+    ] = None
     realtime_session_id: Annotated[str, Field(alias="realtimeSessionId")]
     outcome: ThreadRealtimeSessionOutcome
     type: Annotated[
@@ -11266,16 +11376,22 @@ class RealtimeSessionClosedThreadRealtimeItem(BaseModel):
 class ThreadRealtimeItem(
     RootModel[
         RealtimeSessionStartedThreadRealtimeItem
+        | FrameThreadRealtimeItem
         | TranscriptSegmentThreadRealtimeItem
         | BemItemPromotedThreadRealtimeItem
         | RealtimeSessionClosedThreadRealtimeItem
+        | TurnInputAcceptedThreadRealtimeItem
+        | SessionEndInputSavedThreadRealtimeItem
     ]
 ):
     root: Annotated[
         RealtimeSessionStartedThreadRealtimeItem
+        | FrameThreadRealtimeItem
         | TranscriptSegmentThreadRealtimeItem
         | BemItemPromotedThreadRealtimeItem
-        | RealtimeSessionClosedThreadRealtimeItem,
+        | RealtimeSessionClosedThreadRealtimeItem
+        | TurnInputAcceptedThreadRealtimeItem
+        | SessionEndInputSavedThreadRealtimeItem,
         Field(
             description="EXPERIMENTAL - a thread-scoped realtime item in the canonical timeline."
         ),