Skip to content

armgit / dify

Public Imported from GitHub
0

Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.

6.1 KB
Raw Blame
from collections.abc import Callable
from unittest.mock import MagicMock, create_autospec

import pytest

from machinery.context import RequestContext
from services.knowledge.dataset_access import AccessibleDataset, DatasetAccess, DatasetAccessDeniedError
from services.knowledge.datasets.application import (
    DatasetApplicationService,
    DatasetListFilter,
    DatasetOperations,
    DatasetVisibility,
)
from services.knowledge.datasets.retrieval import retrieval_methods

CONTEXT = RequestContext("request", None, "actor", "tenant")


@pytest.fixture
def operations() -> MagicMock:
    return create_autospec(DatasetOperations, instance=True, spec_set=True)


@pytest.fixture
def access() -> MagicMock:
    result = create_autospec(DatasetAccess, instance=True, spec_set=True)
    result.require_accessible.return_value = AccessibleDataset("dataset", "tenant")
    return result


@pytest.fixture
def service(operations: MagicMock, access: MagicMock) -> DatasetApplicationService:
    return DatasetApplicationService(
        dataset_access=access,
        operations=operations,
        rbac_enabled=True,
        service_api_url="",
        vector_store="milvus",
        tidb_fulltext=False,
    )


@pytest.mark.parametrize(
    ("defaults", "workspace", "overrides"),
    [
        (["dataset.preview"], [], {}),
        ([], ["dataset.full_access", "dataset.create_and_management"], {}),
        ([], [], {"hidden": ["dataset.preview"]}),
    ],
)
def test_restricted_whitelist_takes_precedence_over_grants(
    defaults: list[str], workspace: list[str], overrides: dict[str, list[str]]
) -> None:
    visibility = DatasetVisibility(defaults, workspace, overrides, False, ["allowed"])
    assert visibility.list_scope(rbac_enabled=True) == ["allowed"]


@pytest.mark.parametrize(
    ("visibility", "expected"),
    [
        (DatasetVisibility(default_permissions=["dataset.preview"]), None),
        (DatasetVisibility(workspace_permissions=["dataset.acl.preview"]), None),
        (DatasetVisibility(overrides={"a": ["dataset.preview"], "b": ["dataset.edit"]}), ["a"]),
        (DatasetVisibility(workspace_permissions=["dataset.create_and_management"]), []),
        (DatasetVisibility(unrestricted=False), []),
    ],
)
def test_visibility_scope(visibility: DatasetVisibility, expected: list[str] | None) -> None:
    assert visibility.list_scope(rbac_enabled=True) == expected
    assert visibility.list_scope(rbac_enabled=False) is None


def test_list_applies_visibility_and_reports_model_availability(
    service: DatasetApplicationService, operations: MagicMock
) -> None:
    operations.visibility.return_value = DatasetVisibility(
        default_permissions=["dataset.preview"],
        overrides={"a": ["dataset.edit"]},
        unrestricted=False,
        whitelist_ids=["a"],
    )
    operations.list_datasets.return_value = {
        "data": [
            {
                "id": "a",
                "indexing_technique": "high_quality",
                "embedding_model": "missing",
                "embedding_model_provider": "provider",
            },
            {"id": "b", "indexing_technique": "economy", "embedding_model": None, "embedding_model_provider": None},
        ]
    }
    operations.embedding_models.return_value = set()
    result = service.list_datasets(CONTEXT, DatasetListFilter(ids=["a"]))
    operations.list_datasets.assert_called_once_with(CONTEXT, DatasetListFilter(ids=["a"]), ["a"])
    assert result["data"][0]["embedding_available"] is False
    assert result["data"][0]["permission_keys"] == ["dataset.edit"]
    assert result["data"][1]["embedding_available"] is True


@pytest.mark.parametrize(
    ("method", "extra"),
    [
        (DatasetApplicationService.get_dataset, {}),
        (DatasetApplicationService.update_dataset, {"values": {}}),
        (DatasetApplicationService.delete_dataset, {}),
        (DatasetApplicationService.is_in_use, {}),
        (DatasetApplicationService.queries, {"page": 1, "limit": 20}),
        (DatasetApplicationService.related_apps, {}),
        (DatasetApplicationService.indexing_status, {}),
        (DatasetApplicationService.error_documents, {}),
        (DatasetApplicationService.partial_members, {}),
        (DatasetApplicationService.auto_disable_logs, {}),
        (DatasetApplicationService.set_api_enabled, {"status": "enable"}),
    ],
)
def test_access_denial_precedes_all_owned_operations(
    service: DatasetApplicationService,
    access: MagicMock,
    operations: MagicMock,
    method: Callable[..., object],
    extra: dict[str, object],
) -> None:
    access.require_accessible.side_effect = DatasetAccessDeniedError()
    with pytest.raises(DatasetAccessDeniedError):
        method(service, CONTEXT, dataset_id="dataset", **extra)
    assert operations.mock_calls == []


def test_rbac_create_normalizes_permission_without_mutating_input(
    service: DatasetApplicationService, operations: MagicMock
) -> None:
    values = {"name": "name", "permission": "only_me"}
    service.create_dataset(CONTEXT, values=values)
    operations.create_dataset.assert_called_once_with(CONTEXT, {"name": "name", "permission": "all_team_members"})
    assert values["permission"] == "only_me"


def test_settings_are_injected_and_url_is_normalized(service: DatasetApplicationService) -> None:
    assert service.api_base_url(CONTEXT, request_base_url="https://api.example/v1/") == "https://api.example/v1"
    assert service.retrieval_settings(CONTEXT)["retrieval_method"] == [
        "semantic_search",
        "full_text_search",
        "hybrid_search",
    ]
    assert service.retrieval_settings(CONTEXT, vector_type="milvus", is_mock=True) == {
        "retrieval_method": ["semantic_search"]
    }


@pytest.mark.parametrize("fulltext", [True, False])
def test_tidb_retrieval_setting(fulltext: bool) -> None:
    result = retrieval_methods("tidb_vector", tidb_fulltext=fulltext)
    assert ("full_text_search" in result["retrieval_method"]) is fulltext


@pytest.mark.parametrize("store", [None, "unsupported"])
def test_invalid_vector_store(store: str | None) -> None:
    with pytest.raises(
        ValueError, match=("(?:Unsupported vector db type unsupported\\.|Vector store type is not configured\\.)")
    ):
        retrieval_methods(store)