fix(vector-sync): address PR review — dict guard, symmetric backstop, metrics

- purge route: 400 (not 500) on a valid-JSON non-object body
- scanner: backstop-purge admin-disabled note/news_item/deck_card points
  (their deletion-tracking lives inside the skipped scan_* fns), mirroring the
  files path; gated on a concrete allow-set so fail-open never deletes
- processor: record_ingest_dropped("admin_disabled") so consent-skipped index
  tasks are observable/alertable
- app.py: list /api/v1/vector-sync/purge in the endpoints log line
- capabilities: drop empty-string doc types; return frozenset throughout
- purge: document the count-before-delete approximation
- tests: non-object body -> 400, ProvisioningRequiredError -> 428, cache TTL
  expiry refetch, and the scanner consent backstop

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
Chris Coutinho
2026-06-16 00:54:50 +02:00
co-authored by Claude Opus 4.8
parent ef5b3f3873
commit 477fb02b0a
10 changed files with 258 additions and 43 deletions
+7 -6
View File
@@ -20,10 +20,12 @@ def _patch_qdrant(monkeypatch, *, counts: dict[str, int], delete_raises=None):
def _doc_type_of(flt):
return flt.must[0].match.value
async def fake_count(*, collection_name, count_filter, exact):
# Sync side_effects: AsyncMock awaits the call and returns the value, so the
# helpers don't need to be coroutines themselves.
def fake_count(*, collection_name, count_filter, exact):
return SimpleNamespace(count=counts.get(_doc_type_of(count_filter), 0))
async def fake_delete(*, collection_name, points_selector):
def fake_delete(*, collection_name, points_selector):
dt = _doc_type_of(points_selector)
if delete_raises and dt in delete_raises:
raise RuntimeError(f"delete failed for {dt}")
@@ -31,10 +33,9 @@ def _patch_qdrant(monkeypatch, *, counts: dict[str, int], delete_raises=None):
client.count.side_effect = fake_count
client.delete.side_effect = fake_delete
async def fake_get_qdrant_client():
return client
monkeypatch.setattr(purge_module, "get_qdrant_client", fake_get_qdrant_client)
monkeypatch.setattr(
purge_module, "get_qdrant_client", AsyncMock(return_value=client)
)
monkeypatch.setattr(
purge_module,
"get_settings",
@@ -0,0 +1,81 @@
"""Unit tests for the scanner's admin-consent backstop deletion.
When an admin disables a text source (note/news_item/deck_card), the scanner
skips its scan_* function, so the in-function deletion-tracking never runs. The
backstop enqueues deletes for any indexed points of the disabled type, mirroring
the files path — but only on a concrete allow-set (never on fail-open None).
"""
from __future__ import annotations
from types import SimpleNamespace
from typing import cast
from unittest.mock import AsyncMock
from nextcloud_mcp_server.vector import scanner as scanner_module
from nextcloud_mcp_server.vector.queue.ports import TaskProducer
from nextcloud_mcp_server.vector.scanner import _enqueue_deletes_for_disabled_types
def _producer(send: AsyncMock) -> TaskProducer:
"""A minimal stand-in for the TaskProducer protocol (only ``send`` is used)."""
return cast(TaskProducer, SimpleNamespace(send=send))
def _patch_qdrant(monkeypatch, points_by_type: dict[str, list[str]]):
client = AsyncMock()
def fake_scroll(
*, collection_name, scroll_filter, with_payload, with_vectors, limit, offset
):
# must=[user_id, doc_type] — doc_type is the second condition.
doc_type = scroll_filter.must[1].match.value
points = [
SimpleNamespace(payload={"doc_id": doc_id})
for doc_id in points_by_type.get(doc_type, [])
]
return (points, None)
client.scroll.side_effect = fake_scroll
monkeypatch.setattr(
scanner_module, "get_qdrant_client", AsyncMock(return_value=client)
)
monkeypatch.setattr(
scanner_module,
"get_settings",
lambda: SimpleNamespace(get_collection_name=lambda: "c"),
)
async def test_enqueues_deletes_for_disabled_text_type(monkeypatch):
_patch_qdrant(monkeypatch, {"note": ["n1", "n2"], "deck_card": ["d1"]})
sent: list = []
stream = _producer(AsyncMock(side_effect=lambda t: sent.append(t)))
# note disabled; news_item + deck_card still allowed.
allowed = frozenset({"file", "news_item", "deck_card"})
queued = await _enqueue_deletes_for_disabled_types("alice", stream, allowed, 1)
assert queued == 2
assert {t.doc_id for t in sent} == {"n1", "n2"}
assert all(t.operation == "delete" and t.doc_type == "note" for t in sent)
async def test_noop_when_allowed_is_none(monkeypatch):
# Fail-open: a transient capability read must never trigger deletion.
send = AsyncMock()
queued = await _enqueue_deletes_for_disabled_types(
"alice", _producer(send), None, 1
)
assert queued == 0
send.assert_not_called()
async def test_noop_when_all_text_types_allowed(monkeypatch):
send = AsyncMock()
allowed = frozenset({"note", "news_item", "deck_card", "file"})
queued = await _enqueue_deletes_for_disabled_types(
"alice", _producer(send), allowed, 1
)
assert queued == 0
send.assert_not_called()