~95% on SimpleQA (e.g. Qwen3.6-27B on a 3090). Supports all local and cloud LLMs (llama.cpp, Ollama, Google, ...). 10+ search engines - arXiv, PubMed, your private documents. Everything Local & Encrypted.
Python academiaanthropicarxivbravedeep-researchencryptionhome-automationhomeserverlocallocal-deep-researchlocal-llmmistralollamaopenaipubmedresearchresearch-toolretrieval-augmented-generationsearxngself-hosted
"""Regression coverage for shared rate-limit storage isolation."""
from inspect import unwrap
import pytest
from limits import parse
from local_deep_research.web.dependencies import rate_limit
from tests import conftest as shared_fixtures
@pytest.mark.parametrize("enabled", [True, False], ids=["enabled", "disabled"])
def test_shared_fixture_resets_storage_without_changing_enforcement(
monkeypatch: pytest.MonkeyPatch, enabled: bool
) -> None:
"""Both fixture phases clear storage without changing enforcement settings."""
limiter = rate_limit.limiter
monkeypatch.setattr(limiter, "enabled", enabled)
rate = parse("1 per hour")
scope = "route-fixture-reset-regression"
limiter.reset()
reset_fixture = unwrap(shared_fixtures.reset_all_singletons)()
try:
assert limiter._limiter.hit(rate, scope)
assert not limiter._limiter.hit(rate, scope)
next(reset_fixture)
assert limiter._limiter.hit(rate, scope)
assert not limiter._limiter.hit(rate, scope)
assert limiter.enabled is enabled
with pytest.raises(StopIteration):
next(reset_fixture)
assert limiter._limiter.hit(rate, scope)
assert not limiter._limiter.hit(rate, scope)
assert limiter.enabled is enabled
finally:
reset_fixture.close()
limiter.reset()