Testing (quadkit-testing)

Test harnesses, fakes, and fixtures for Quadkit applications: boot the real application in-process, substitute bindings instead of mocking import sites, and assert on responses with helpers that say what failed.

For anyone writing tests against Quadkit applications — and for tooling that needs drop-in implementations of the framework’s protocols.

The quadkit family

| Package | Role | | --- | --- | | quadkit-contracts | zero-dependency protocols, types, exception hierarchy | | quadkit | the framework core — DI container, modules, config, logging, Result | | quadkit-web | ASGI layer — controllers, routing, middleware, OpenAPI docs | | quadkit-cli | project scaffolding and code generators | | quadkit-testing | in-process test beds, fakes, fixtures |

Installation

uv add --dev quadkit-testing

Requires Python >= 3.11.

Minimal working example

Test an HTTP route with the real application, in-process:

import pytest

from quadkit.testing import WebTestBed

from my_app import create_app


@pytest.mark.asyncio
async def test_hello() -> None:
    async with WebTestBed(create_app()) as bed:
        response = bed.get("/hello", params={"name": "quadkit"})
        response.assert_status(200)
        assert response.json == {"message": "hello, quadkit"}

Or test services directly, with binding overrides:

async def test_service() -> None:
    from quadkit.testing import AppTestBed

    async with AppTestBed.from_factory(
        create_app, overrides={Cache: FakeCache()}
    ) as bed:
        service = await bed.app.container.resolve(UserService)

A pytest plugin registers automatically via entry points — no conftest.py wiring — and provides auto-registered fixtures including fake_cache, fake_event_bus, fake_logger, fake_clock, fake_command_bus, fake_query_bus, fake_unit_of_work, fake_metrics, fake_config, fake_state_store, test_bed, test_container, and test_data:

import pytest

from quadkit.contracts.domain.events import DomainEvent


class UserCreated(DomainEvent):
    user_id: str
    email: str


@pytest.mark.asyncio
async def test_signup_publishes(fake_event_bus) -> None:
    await fake_event_bus.publish(UserCreated(user_id="1", email="[email protected]"))
    fake_event_bus.assert_published(UserCreated, user_id="1")


def test_trial_expiry_is_deterministic(fake_clock) -> None:
    t0 = fake_clock.now()
    fake_clock.advance(30 * 24 * 3600)
    assert (fake_clock.now() - t0).days == 30

Optional extras

| Extra | Contents | | --- | --- | | [web] | httpx + Starlette — the WebTestBed client transport | | [db] | aiosqlite, asyncpg — drivers for your async DB test suites | | [integration] | service clients for integration suites (Redis, MongoDB, Kafka, Elasticsearch, Neo4j, Qdrant, PostgreSQL, SQLite) | | [dev] | ruff, mypy, black |

Public API entry points

Test beds

from quadkit.testing import AppTestBed, WebTestBed

AppTestBed.from_factory(factory, overrides=None) / AppTestBed.from_app(app) — application-level beds.
WebTestBed(app_or_provider, raise_server_exceptions=True) with get/post/put/patch/delete, override(Contract, impl) (before boot), and TestResponse — status_code, headers, text, json (property), assert_status, assert_json, assert_json_path, assert_header.
quadkit.testing.fixtures.container.ContainerTestFixture — DI-level fixture with mock(), override(), get(), get_optional().
quadkit.testing.fixtures.bed.TestEnvironment — programmatic environment builder (use_provider, override, fake, resolve).
quadkit.testing.lib.factory.TestDataFactory — deterministic create_user(), create_task(), create_message(), create_request().

Fakes (`quadkit.testing.fakes`)

All in-process, async-native, implementing the same contracts as the real services:

| Class | Covers | | --- | --- | | FakeCache / FakeStateStore | cache and state storage | | FakeEventBus | in-process events with assert_published(), published_of_type(), assert_events_in_order() and friends | | FakeCommandBus / FakeQueryBus | command / query dispatch | | FakeUnitOfWork | unit-of-work context | | FakeClock (+ Clock, SystemClock) | deterministic time | | FakeConfig | config overrides | | FakeLogger (+ LogEntry) | structlog-compatible sink | | FakeMetricsCollector / FakeResourceUnitTracker | metrics / resource tracking | | FakeRedisClient | Redis-protocol client | | FakeAuditLogger | audit records | | FakeTracer / FakeSpan | tracing |

The published wheel carries the ai, db and web test clients and beds plus the fakes and fixtures that need no private package. The auth, cache, events, search, storage, tasks, ui test clients, the AI/DB/task fixture modules, the secrets fake and IntegrationEnvironment stay in this repository until their packages publish; asking the published wheel for one raises AttributeError naming the module.

Configuration

None required — the pytest plugin self-registers. Mark suites for external services and gate them yourself (e.g. uv run pytest -m "not integration").

Error handling

Test beds surface failures, they don’t hide them: with raise_server_exceptions=True (the default) unexpected exceptions re-raise into your test with their original traceback; HTTP-expected failures assert on the response instead.

Testing

Ironically self-hosted: this package’s public tests are among the suite the release executes from the exported tree against the built wheels.

Security

Fakes are in-process and safe to wire into unit suites. Report vulnerabilities privately per SECURITY.md.

Stability

Version 0.0.42 in the 0.x series, released in lockstep with the other four distributions; APIs may change between minor versions until 1.0 — pin an exact version (quadkit-testing==0.0.42) or a tight range (>=0.0.42,<0.1.0). Full policy: stability and compatibility.