From 43bcdcf97f9ece5151ff334333791028b0c17c84 Mon Sep 17 00:00:00 2001
From: ruizanthony <ruizanthony@gmail.com>
Date: Thu, 20 Aug 2026 03:01:10 +0200
Subject: [PATCH] fix(transcript): recognize Agent _row_id during replay merge
 (#7133)

* fix(transcript): recognize agent row ids in replay merge

* Strip imported _row_id so replay merge cannot trust caller provenance.

Session import already dropped _state_db_row_id aliases. After treating
Agent _row_id as identity, the same public/import scrubber must drop that
field too, or a crafted import can collide with a later state.db turn.

---------

Co-authored-by: Anthony Ruiz <ruizanthony@users.noreply.github.com>
Co-authored-by: nesquena-hermes <nesquena+hermes@gmail.com>
---
 api/helpers.py                          |  3 +-
 api/models.py                           |  2 +-
 tests/test_agent_row_id_replay_order.py | 90 +++++++++++++++++++++++++
 3 files changed, 93 insertions(+), 2 deletions(-)
 create mode 100644 tests/test_agent_row_id_replay_order.py

diff --git a/api/helpers.py b/api/helpers.py
index 7349cccb..780f5915 100644
--- a/api/helpers.py
+++ b/api/helpers.py
@@ -16,6 +16,7 @@ logger = logging.getLogger(__name__)
 
 _PUBLIC_MESSAGE_INTERNAL_FIELDS = frozenset({
     "api_content",
+    "_row_id",
     "_state_db_row_id",
     "_db_row_id",
     "state_db_row_id",
@@ -1194,7 +1195,7 @@ def strip_public_internal_fields(value, *, message_records: bool = False):
     """Deep-copy imported records through the shared schema scrubber.
 
     JSON import uses this before constructing or saving a ``Session``.  The
-    Four replay aliases belong to a message/content-part/tool-call/function
+    Five replay aliases belong to a message/content-part/tool-call/function
     record itself; matching names inside user content or tool arguments are
     ordinary JSON and must be preserved.  This is intentionally independent of
     the credential-redaction setting: caller-supplied provider sidecars must
diff --git a/api/models.py b/api/models.py
index 60df8c7b..3ceca910 100644
--- a/api/models.py
+++ b/api/models.py
@@ -8646,7 +8646,7 @@ def _state_db_row_identity_details(message: dict | None) -> tuple[str | None, bo
     if not isinstance(message, dict):
         return None, True
     values = set()
-    for key in ("_state_db_row_id", "_db_row_id", "state_db_row_id"):
+    for key in ("_row_id", "_state_db_row_id", "_db_row_id", "state_db_row_id"):
         if key not in message or message.get(key) in (None, ""):
             continue
         value = message.get(key)
diff --git a/tests/test_agent_row_id_replay_order.py b/tests/test_agent_row_id_replay_order.py
new file mode 100644
index 00000000..22acf521
--- /dev/null
+++ b/tests/test_agent_row_id_replay_order.py
@@ -0,0 +1,90 @@
+"""Regression: Agent ``_row_id`` must reconcile with WebUI state provenance."""
+
+
+def test_agent_and_webui_row_id_aliases_resolve_to_the_same_sqlite_row():
+    from api.models import _state_db_row_identity_details
+
+    agent_identity, agent_valid = _state_db_row_identity_details({"_row_id": 780107})
+    webui_identity, webui_valid = _state_db_row_identity_details(
+        {"_state_db_row_id": 780107}
+    )
+
+    assert agent_valid is True
+    assert webui_valid is True
+    assert agent_identity == webui_identity
+    assert agent_identity is not None
+
+
+def test_agent_row_id_alias_dedupes_restamped_prompt_before_final_answer():
+    from api.models import merge_session_messages_append_only
+
+    sidecar_messages = [
+        {
+            "role": "assistant",
+            "content": "earlier activity",
+            "timestamp": 100.0,
+            "_state_db_row_id": 700,
+            "api_content": "earlier activity wire",
+        },
+        {
+            "role": "user",
+            "content": "original maintenance question",
+            "timestamp": 110.0,
+            "_row_id": 780107,
+            "api_content": "original maintenance question wire",
+        },
+        {
+            "role": "assistant",
+            "content": "settled final answer",
+            "timestamp": 130.0,
+        },
+    ]
+    state_messages = [
+        {
+            "role": "user",
+            "content": "original maintenance question",
+            "timestamp": 120.0,
+            "_state_db_row_id": 780107,
+            "api_content": "original maintenance question wire",
+        }
+    ]
+
+    merged = merge_session_messages_append_only(sidecar_messages, state_messages)
+
+    assert [message["content"] for message in merged] == [
+        "earlier activity",
+        "original maintenance question",
+        "settled final answer",
+    ]
+
+
+def test_imported_row_id_is_not_trusted_as_state_identity():
+    """JSON import must not keep a caller-supplied ``_row_id`` as provenance.
+
+    Replay merge now treats ``_row_id`` as an Agent/WebUI identity alias. Import
+    must strip that field the same way it strips ``_state_db_row_id``, or a
+    crafted import can collide with a later genuine state.db turn.
+    """
+    from api.helpers import strip_public_internal_fields
+    from api.models import _state_db_row_identity_details
+
+    imported = strip_public_internal_fields(
+        {
+            "messages": [
+                {
+                    "role": "user",
+                    "content": "same prompt",
+                    "timestamp": 10.0,
+                    "_row_id": 780107,
+                    "_state_db_row_id": 780107,
+                }
+            ]
+        }
+    )
+    imported_msg = imported["messages"][0]
+    assert isinstance(imported_msg, dict)
+    assert "_row_id" not in imported_msg
+    assert "_state_db_row_id" not in imported_msg
+    identity, valid = _state_db_row_identity_details(imported_msg)
+    assert valid is True
+    assert identity is None