LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows
Python ai-agentsdata-engineeringdocument-aigenerative-aiidpjson-extractionllmmcp-serverocrpdf-extractionprompt-engineeringstructured-output
"""Contract tests for utils.pagination.OptionalPagination.
The four listing endpoints it backs (workflows, prompt studio, adapters,
connectors) are shared with dropdown/selector consumers that expect a bare
array. The opt-in behaviour below is what keeps those consumers regression-free:
no page/page_size param -> paginate_queryset returns None -> DRF serialises the
bare list; only an explicit opt-in gets the {count, next, previous, results}
envelope.
Deliberately DB-free and Django-settings-free (a hand-rolled request stub, no
APIRequestFactory) so it runs in the rig's unit tier.
"""
from __future__ import annotations
from utils.pagination import OptionalPagination
QUERYSET = list(range(1, 101)) # 100 sliceable items stand in for a queryset
class _Req:
"""Minimal stand-in exposing the surface DRF pagination touches."""
def __init__(self, params: dict[str, str]):
self.query_params = params
def build_absolute_uri(self) -> str:
return "https://testserver/things/"
class TestOptionalPagination:
def test_no_params_returns_none(self):
"""No opt-in -> None, so callers keep their bare-array response."""
assert OptionalPagination().paginate_queryset(QUERYSET, _Req({})) is None
def test_unrelated_params_do_not_trigger_pagination(self):
req = _Req({"adapter_type": "LLM", "search": "foo"})
assert OptionalPagination().paginate_queryset(QUERYSET, req) is None
def test_blank_params_do_not_trigger_pagination(self):
"""Empty ?page= / ?page_size= keep the bare-array response."""
req = _Req({"page": "", "page_size": ""})
assert OptionalPagination().paginate_queryset(QUERYSET, req) is None
def test_single_blank_param_does_not_trigger_pagination(self):
for params in ({"page": ""}, {"page_size": ""}):
assert OptionalPagination().paginate_queryset(QUERYSET, _Req(params)) is None
def test_blank_page_with_explicit_page_size_serves_first_page(self):
"""A blank ?page= alongside a real ?page_size= is page 1, not an error.
DRF reads the page number as `query_params.get("page") or 1`, so a blank
value falls back to the first page instead of raising NotFound.
"""
page = OptionalPagination().paginate_queryset(
QUERYSET, _Req({"page": "", "page_size": "10"})
)
assert page == list(range(1, 11))
def test_blank_page_size_with_explicit_page_uses_default_size(self):
page = OptionalPagination().paginate_queryset(
QUERYSET, _Req({"page": "2", "page_size": ""})
)
assert page == list(range(51, 101))
def test_page_param_opts_in(self):
paginator = OptionalPagination()
page = paginator.paginate_queryset(QUERYSET, _Req({"page": "1"}))
assert page == list(range(1, 51)) # default page_size == 50
assert paginator.page.paginator.count == 100
def test_page_size_param_alone_opts_in(self):
page = OptionalPagination().paginate_queryset(QUERYSET, _Req({"page_size": "10"}))
assert page == list(range(1, 11))
def test_page_size_capped_at_max(self):
page = OptionalPagination().paginate_queryset(
QUERYSET, _Req({"page_size": "100000"})
)
# max_page_size (1000) caps the request; all 100 rows fit in one page
assert page == QUERYSET