Skip to content

Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.

fix(vectordb): enforce the UINT16 length contract for list<string> elements (#3681)

bianbiandashenef577033b581
diff --git a/openviking/storage/vectordb/store/bytes_row.py b/openviking/storage/vectordb/store/bytes_row.py
index cd9f2ad00..73011924e 100644
--- a/openviking/storage/vectordb/store/bytes_row.py
+++ b/openviking/storage/vectordb/store/bytes_row.py
@@ -213,6 +213,10 @@ class _PyBytesRow:
for item in value:
bytes_item = item.encode("utf-8")
bytes_item_len = len(bytes_item)
+ if bytes_item_len > STRING_MAX_UINT16_LENGTH:
+ raise ValueError(
+ f"string element in list field '{field_meta.name}' exceeds 65535 bytes"
+ )
var_fmt_list.append("H")
var_val_list.append(bytes_item_len)
var_fmt_list.append(f"{bytes_item_len}s")
diff --git a/tests/vectordb/test_bytes_row.py b/tests/vectordb/test_bytes_row.py
index 394dcccae..8b197e2f1 100644
--- a/tests/vectordb/test_bytes_row.py
+++ b/tests/vectordb/test_bytes_row.py
@@ -125,6 +125,22 @@ class TestBytesRow(unittest.TestCase):
with self.assertRaisesRegex(ValueError, "text.*exceeds 65535 bytes"):
py_row.serialize({"text": text})
+ def test_oversized_list_string_element_serialization_raises(self):
+ # A list<string> element uses the same UINT16 length prefix as a scalar
+ # string, so it must reject >65535-byte elements with the same clean,
+ # field-attributed ValueError — not a raw struct.error from deep inside
+ # struct.pack_into.
+ py_schema = _PySchema(
+ [{"name": "tags", "data_type": _PyFieldType.list_string, "id": 0}]
+ )
+ py_row = _PyBytesRow(py_schema)
+ with self.assertRaisesRegex(ValueError, "tags.*exceeds 65535 bytes"):
+ py_row.serialize({"tags": ["x" * 65536, "ok"]})
+
+ # A list of in-bounds strings (including multibyte) still round-trips.
+ blob = py_row.serialize({"tags": ["hello", "世界"]})
+ self.assertEqual(py_row.deserialize_field(blob, "tags"), ["hello", "世界"])
+
def test_binary_data(self):
@serializable
@dataclass