Skip to content

RAGFlow is an agent-native database: a Retrieval-Augmented Generation (RAG) engine that grew into context infrastructure.

fix: apply order_by return value in deep-pagination helpers (unordered OFFSET/LIMIT can skip/dupe rows) (#17325)

rsnetworkinginc00674709033c
diff --git a/api/db/services/canvas_service.py b/api/db/services/canvas_service.py
index d5a3529e5b..07b7025ef3 100644
--- a/api/db/services/canvas_service.py
+++ b/api/db/services/canvas_service.py
@@ -74,7 +74,7 @@ class UserCanvasService(CommonService):
# find team agents and owned agents
agents = cls.model.select(*fields).where((cls.model.user_id.in_(tenant_ids) & (cls.model.permission == TenantPermission.TEAM.value)) | (cls.model.user_id == user_id))
# sort by create_time, asc
- agents.order_by(cls.model.create_time.asc())
+ agents = agents.order_by(cls.model.create_time.asc())
# maybe cause slow query by deep paginate, optimize later
offset, limit = 0, 50
res = []
diff --git a/api/db/services/conversation_service.py b/api/db/services/conversation_service.py
index 4bfd44b80b..60765cf184 100644
--- a/api/db/services/conversation_service.py
+++ b/api/db/services/conversation_service.py
@@ -172,7 +172,7 @@ class ConversationService(CommonService):
@DB.connection_context()
def get_all_conversation_by_dialog_ids(cls, dialog_ids):
sessions = cls.model.select().where(cls.model.dialog_id.in_(dialog_ids))
- sessions.order_by(cls.model.create_time.asc())
+ sessions = sessions.order_by(cls.model.create_time.asc())
offset, limit = 0, 100
res = []
while True:
diff --git a/api/db/services/dialog_service.py b/api/db/services/dialog_service.py
index 12f14dd020..cccd19f752 100644
--- a/api/db/services/dialog_service.py
+++ b/api/db/services/dialog_service.py
@@ -263,7 +263,7 @@ class DialogService(CommonService):
def get_all_dialogs_by_tenant_id(cls, tenant_id):
fields = [cls.model.id]
dialogs = cls.model.select(*fields).where(cls.model.tenant_id == tenant_id)
- dialogs.order_by(cls.model.create_time.asc())
+ dialogs = dialogs.order_by(cls.model.create_time.asc())
offset, limit = 0, 100
res = []
while True:
diff --git a/api/db/services/document_service.py b/api/db/services/document_service.py
index 67a32b0615..b5f866fc34 100644
--- a/api/db/services/document_service.py
+++ b/api/db/services/document_service.py
@@ -362,7 +362,7 @@ class DocumentService(CommonService):
def get_all_doc_ids_by_kb_ids(cls, kb_ids):
fields = [cls.model.id, cls.model.kb_id]
docs = cls.model.select(*fields).where(cls.model.kb_id.in_(kb_ids))
- docs.order_by(cls.model.create_time.asc())
+ docs = docs.order_by(cls.model.create_time.asc())
# maybe cause slow query by deep paginate, optimize later
offset, limit = 0, 100
res = []
@@ -436,7 +436,7 @@ class DocumentService(CommonService):
def get_all_docs_by_creator_id(cls, creator_id):
fields = [cls.model.id, cls.model.kb_id, cls.model.token_num, cls.model.chunk_num, Knowledgebase.tenant_id]
docs = cls.model.select(*fields).join(Knowledgebase, on=(Knowledgebase.id == cls.model.kb_id)).where(cls.model.created_by == creator_id)
- docs.order_by(cls.model.create_time.asc())
+ docs = docs.order_by(cls.model.create_time.asc())
# maybe cause slow query by deep paginate, optimize later
offset, limit = 0, 100
res = []
diff --git a/api/db/services/file_service.py b/api/db/services/file_service.py
index 6ccd57b523..5af5b04f1c 100644
--- a/api/db/services/file_service.py
+++ b/api/db/services/file_service.py
@@ -215,7 +215,7 @@ class FileService(CommonService):
def get_all_file_ids_by_tenant_id(cls, tenant_id):
fields = [cls.model.id]
files = cls.model.select(*fields).where(cls.model.tenant_id == tenant_id)
- files.order_by(cls.model.create_time.asc())
+ files = files.order_by(cls.model.create_time.asc())
offset, limit = 0, 100
res = []
while True:
diff --git a/api/db/services/knowledgebase_service.py b/api/db/services/knowledgebase_service.py
index 2d906ad1e4..8b8556c3e1 100644
--- a/api/db/services/knowledgebase_service.py
+++ b/api/db/services/knowledgebase_service.py
@@ -289,7 +289,7 @@ class KnowledgebaseService(CommonService):
# find team kb and owned kb
kbs = cls.model.select(*fields).where(cls._visibility_and_status_filter(tenant_ids, user_id))
# sort by create_time asc
- kbs.order_by(cls.model.create_time.asc())
+ kbs = kbs.order_by(cls.model.create_time.asc())
# maybe cause slow query by deep paginate, optimize later.
offset, limit = 0, 50
res = []
diff --git a/api/db/services/user_canvas_version.py b/api/db/services/user_canvas_version.py
index 315312f572..7fe4434d5c 100644
--- a/api/db/services/user_canvas_version.py
+++ b/api/db/services/user_canvas_version.py
@@ -55,7 +55,7 @@ class UserCanvasVersionService(CommonService):
def get_all_canvas_version_by_canvas_ids(cls, canvas_ids):
fields = [cls.model.id]
versions = cls.model.select(*fields).where(cls.model.user_canvas_id.in_(canvas_ids))
- versions.order_by(cls.model.create_time.asc())
+ versions = versions.order_by(cls.model.create_time.asc())
offset, limit = 0, 100
res = []
while True:
diff --git a/test/unit_test/api/db/services/test_deep_pagination_order_by.py b/test/unit_test/api/db/services/test_deep_pagination_order_by.py
new file mode 100644
index 0000000000..4ce5a90a09
--- /dev/null
+++ b/test/unit_test/api/db/services/test_deep_pagination_order_by.py
@@ -0,0 +1,140 @@
+#
+# Copyright 2026 The InfiniFlow Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+import warnings
+from types import SimpleNamespace
+
+import pytest
+
+warnings.filterwarnings(
+ "ignore",
+ message="pkg_resources is deprecated as an API.*",
+ category=UserWarning,
+)
+
+from api.db.services import conversation_service, dialog_service, file_service, user_canvas_version
+
+
+class _FakeOrderField:
+ def asc(self):
+ return ("create_time", "asc")
+
+ def desc(self):
+ return ("create_time", "desc")
+
+
+class _FakeField:
+ def __eq__(self, other):
+ return ("eq", other)
+
+ def in_(self, other):
+ return ("in", other)
+
+
+class _FakeQuery:
+ """Mimic peewee's immutable query chaining.
+
+ peewee query methods (``where``/``order_by``/``offset``/``limit``) return a
+ *new* query object and never mutate in place. Modelling that faithfully is
+ what lets these tests catch the "``order_by()`` result discarded" bug: the
+ deep-pagination helpers previously called ``q.order_by(...)`` on its own
+ line, throwing the ordered query away and paginating an unordered one.
+ """
+
+ def __init__(self, rows, ordered=False, offset=0, limit=None):
+ self._rows = rows
+ self._ordered = ordered
+ self._offset = offset
+ self._limit = limit
+
+ def where(self, *args, **kwargs):
+ return _FakeQuery(self._rows, self._ordered, self._offset, self._limit)
+
+ def order_by(self, *args, **kwargs):
+ return _FakeQuery(self._rows, True, self._offset, self._limit)
+
+ def offset(self, offset):
+ return _FakeQuery(self._rows, self._ordered, offset, self._limit)
+
+ def limit(self, limit):
+ return _FakeQuery(self._rows, self._ordered, self._offset, limit)
+
+ def dicts(self):
+ rows = sorted(self._rows, key=lambda r: r["create_time"]) if self._ordered else list(self._rows)
+ if self._limit is not None:
+ rows = rows[self._offset : self._offset + self._limit]
+ return rows
+
+
+# ``create_time`` is deliberately out of insertion order so that a missing
+# ``ORDER BY create_time`` is observable in the returned sequence.
+_ROWS = [
+ {"id": "c", "create_time": 3},
+ {"id": "a", "create_time": 1},
+ {"id": "b", "create_time": 2},
+]
+
+
+def _fake_model():
+ return SimpleNamespace(
+ select=lambda *args, **kwargs: _FakeQuery(list(_ROWS)),
+ id=_FakeField(),
+ dialog_id=_FakeField(),
+ tenant_id=_FakeField(),
+ user_canvas_id=_FakeField(),
+ create_time=_FakeOrderField(),
+ )
+
+
+def _patch(monkeypatch, module, service):
+ monkeypatch.setattr(module.DB, "connect", lambda *args, **kwargs: None)
+ monkeypatch.setattr(module.DB, "close", lambda *args, **kwargs: None)
+ monkeypatch.setattr(service, "model", _fake_model())
+
+
+def test_conversation_helper_orders_by_create_time(monkeypatch):
+ _patch(monkeypatch, conversation_service, conversation_service.ConversationService)
+
+ rows = conversation_service.ConversationService.get_all_conversation_by_dialog_ids(["dialog-1"])
+
+ assert [r["id"] for r in rows] == ["a", "b", "c"]
+
+
+def test_file_helper_orders_by_create_time(monkeypatch):
+ _patch(monkeypatch, file_service, file_service.FileService)
+
+ rows = file_service.FileService.get_all_file_ids_by_tenant_id("tenant-1")
+
+ assert [r["id"] for r in rows] == ["a", "b", "c"]
+
+
+def test_dialog_helper_orders_by_create_time(monkeypatch):
+ _patch(monkeypatch, dialog_service, dialog_service.DialogService)
+
+ rows = dialog_service.DialogService.get_all_dialogs_by_tenant_id("tenant-1")
+
+ assert [r["id"] for r in rows] == ["a", "b", "c"]
+
+
+def test_canvas_version_helper_orders_by_create_time(monkeypatch):
+ _patch(monkeypatch, user_canvas_version, user_canvas_version.UserCanvasVersionService)
+
+ rows = user_canvas_version.UserCanvasVersionService.get_all_canvas_version_by_canvas_ids(["canvas-1"])
+
+ assert [r["id"] for r in rows] == ["a", "b", "c"]