perf(streaming): bound terminal session payloads

api/gateway_chat.py

@@ -34,7 +34,7 @@
     unregister_stream_owner,
     update_active_run,
 )
-from api.helpers import _redact_text, redact_session_data
+from api.helpers import _redact_text
 from api.models import clear_process_wakeup_pause, get_session, merge_session_messages_append_only
 from api.run_journal import RunJournalWriter, bound_run_journal_snapshot_args
 
@@ -737,7 +737,7 @@ def _settle_gateway_terminal_error(session_id, stream_id, workspace, model, mode
         _classify_provider_error,
         _materialize_pending_user_turn_before_error,
         _provider_error_payload,
-        _session_payload_with_full_messages,
+        _best_effort_terminal_session_payload,
         _snapshot_and_append_partial_on_error,
         _terminal_turn_duration,
     )
@@ -788,9 +788,9 @@ def _settle_gateway_terminal_error(session_id, stream_id, workspace, model, mode
             terminal_session_persisted = True
         except Exception:
             logger.debug("Failed to persist gateway terminal error settlement", exc_info=True)
-        error_payload["session"] = redact_session_data(
-            _session_payload_with_full_messages(session, tool_calls=[])
-        )
+        terminal_session_payload = _best_effort_terminal_session_payload(session)
+        if terminal_session_payload is not None:
+            error_payload["session"] = terminal_session_payload
         error_payload["session_id"] = session.session_id
         error_payload["terminal_session_persisted"] = terminal_session_persisted
         if terminal_session_persisted:
@@ -1351,9 +1351,12 @@ def _restore_cancelled_success_writeback():
                 session_id,
                 goal_exc,
             )
-        from api.streaming import _session_payload_with_full_messages
-        gateway_session_payload = _session_payload_with_full_messages(s, tool_calls=[])
-        put_gateway_event("done", {"session": redact_session_data(gateway_session_payload), "usage": usage})
+        from api.streaming import _best_effort_terminal_session_payload
+        gateway_done_payload = {"usage": usage}
+        gateway_session_payload = _best_effort_terminal_session_payload(s)
+        if gateway_session_payload is not None:
+            gateway_done_payload["session"] = gateway_session_payload
+        put_gateway_event("done", gateway_done_payload)
         put_gateway_event("stream_end", {"session_id": session_id})
     except urllib.error.HTTPError as exc:
         try:

api/routes.py

@@ -61,6 +61,7 @@
 )
 from api.gateway_restart import restart_active_profile_gateway
 from api.shares import create_or_refresh_share, load_share, revoke_share
+from api.session_ops import _messages_for_limited_payload
 
 logger = logging.getLogger(__name__)
 
@@ -8772,7 +8773,6 @@ def _message_window_for_display(messages, msg_limit=None, msg_before=None, expan
     return window, start_idx
 
 
-_LIMITED_TOOL_CONTENT_MAX_CHARS = 4096
 # Server-side ceiling on the ?msg_limit= tail-window size. A client could
 # otherwise request msg_limit=1000000 and force the server to assemble and
 # serialize an unbounded message payload (the frontend's own pagination grows
@@ -8841,48 +8841,6 @@ def _state_db_backstop_limit_for_display(session, msg_before) -> int | None:
     return None if has_boundary_prefix else _STATE_DB_DISPLAY_ROW_BACKSTOP
 
 
-_LIMITED_TOOL_CONTENT_NOTICE = (
-    "\n\n[Tool output truncated in paginated session response; "
-    "load the full transcript to inspect the complete result.]"
-)
-
-
-def _tool_message_for_limited_payload(message):
-    """Return a bounded copy of large hidden tool-result rows for paginated loads."""
-    if not isinstance(message, dict) or str(message.get("role") or "").lower() != "tool":
-        return message
-    content = message.get("content")
-    if content in (None, ""):
-        return message
-    if isinstance(content, str):
-        text = content
-    else:
-        try:
-            text = json.dumps(content, ensure_ascii=False, default=str)
-        except Exception:
-            text = str(content)
-    if len(text) <= _LIMITED_TOOL_CONTENT_MAX_CHARS:
-        return message
-    clipped = dict(message)
-    preview = text[:_LIMITED_TOOL_CONTENT_MAX_CHARS] + _LIMITED_TOOL_CONTENT_NOTICE
-    if isinstance(content, str):
-        clipped["content"] = preview
-    elif isinstance(content, list):
-        clipped["content"] = [{"type": "text", "text": preview}]
-    elif isinstance(content, dict):
-        clipped["content"] = {"_truncated": True, "preview": preview}
-    else:
-        clipped["content"] = preview
-    clipped["_content_truncated"] = True
-    clipped["_content_original_chars"] = len(text)
-    return clipped
-
-
-def _messages_for_limited_payload(messages) -> list:
-    """Bound hidden tool-result payloads before sending a msg_limit response."""
-    return [_tool_message_for_limited_payload(msg) for msg in list(messages or [])]
-
-
 def _limited_webui_messages_for_display(session, state_db_messages) -> list:
     """Return the display sidecar plus only necessary state.db rows for msg_limit.
 

api/session_ops.py

@@ -17,6 +17,47 @@
 logger = logging.getLogger(__name__)
 
 AUTO_TITLE_LABELS = {'untitled', 'new chat'}
+_LIMITED_TOOL_CONTENT_MAX_CHARS = 4096
+_LIMITED_TOOL_CONTENT_NOTICE = (
+    "\n\n[Tool output truncated in paginated session response; "
+    "load the full transcript to inspect the complete result.]"
+)
+
+
+def _tool_message_for_limited_payload(message):
+    """Return a bounded copy of a large hidden tool-result row."""
+    if not isinstance(message, dict) or str(message.get("role") or "").lower() != "tool":
+        return message
+    content = message.get("content")
+    if content in (None, ""):
+        return message
+    if isinstance(content, str):
+        text = content
+    else:
+        try:
+            text = json.dumps(content, ensure_ascii=False, default=str)
+        except Exception:
+            text = str(content)
+    if len(text) <= _LIMITED_TOOL_CONTENT_MAX_CHARS:
+        return message
+    clipped = dict(message)
+    preview = text[:_LIMITED_TOOL_CONTENT_MAX_CHARS] + _LIMITED_TOOL_CONTENT_NOTICE
+    if isinstance(content, str):
+        clipped["content"] = preview
+    elif isinstance(content, list):
+        clipped["content"] = [{"type": "text", "text": preview}]
+    elif isinstance(content, dict):
+        clipped["content"] = {"_truncated": True, "preview": preview}
+    else:
+        clipped["content"] = preview
+    clipped["_content_truncated"] = True
+    clipped["_content_original_chars"] = len(text)
+    return clipped
+
+
+def _messages_for_limited_payload(messages) -> list:
+    """Bound hidden tool-result rows before a limited session payload is sent."""
+    return [_tool_message_for_limited_payload(message) for message in list(messages or [])]
 
 
 def _live_active_stream_id(session) -> str | None: