LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows
Python ai-agentsdata-engineeringdocument-aigenerative-aiidpjson-extractionllmmcp-serverocrpdf-extractionprompt-engineeringstructured-output
"""Top-level pytest conftest for the backend service.
Loads environment variables from ``test.env`` before pytest collects any
test modules. This replaces the ``pytest-dotenv`` plugin, which has been
unreleased since Feb 2020 and is a recurring break risk across pytest
majors.
Behavior matches the previous ``env_files = "test.env"`` setting under
``[tool.pytest.ini_options]``: variables already in the process
environment are preserved (``override=False``); a missing file is
silently tolerated.
Note: backend tests do not run under tox in CI today (no ``backend``
testenv); this file only takes effect in local/IDE runs.
"""
from pathlib import Path
import pytest
from dotenv import load_dotenv
# `-s` is set in pyproject so prints surface — log when test.env is absent
# to make a mis-located file debuggable instead of silently empty.
if not load_dotenv(Path(__file__).parent / "test.env", override=False):
print("[conftest] backend/test.env not found; using ambient environment", flush=True)
@pytest.fixture(autouse=True)
def _drop_cached_organization():
"""Keep the per-request organization cache from crossing tests.
Tests share a thread, and many leave the organization id set. Production
drops the cache when the next request sets the id, or at request end; a
test that relies on a leftover id would instead be served an organization
another test created and rolled back.
"""
yield
# Django-free on purpose: ``utils.user_context`` (which registers the cache
# as a derived key) needs the app registry, and ``utils.constants`` needs
# configured settings; a run of the Django-free test files has neither, and
# then has nothing registered to drop either.
from utils.local_context import StateStore
StateStore.drop_derived()
def pytest_collection_modifyitems(items):
"""Auto-mark every DB-bound test as ``integration`` so the rig's unit tier
(``-m 'not integration'``) skips it while the integration tier (live
Postgres) runs it. Detects Django ``TestCase``/``TransactionTestCase``
subclasses and any item using the ``django_db`` marker — the two ways a
backend test needs a database. Kept central so tests declare their DB need
by how they're written, not by a hand-maintained marker on each file.
"""
from django.test import TestCase, TransactionTestCase
for item in items:
cls = getattr(item, "cls", None)
needs_db = item.get_closest_marker("django_db") is not None or (
cls is not None and issubclass(cls, (TestCase, TransactionTestCase))
)
if needs_db:
item.add_marker(pytest.mark.integration)