~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
"""Tests for the search-engine selector grouping/ordering logic."""
from local_deep_research.web_search_engines.engine_groups import (
FAVORITES_GROUP_KEY,
SEARCH_ENGINE_GROUPS,
classify_engine_group,
effective_group,
group_icon,
group_label,
group_order,
is_collection_engine,
)
class TestIsCollectionEngine:
def test_library_aggregate_is_collection(self):
assert is_collection_engine("library") is True
def test_per_collection_pseudo_engine(self):
assert is_collection_engine("collection_42") is True
def test_regular_engine_is_not_collection(self):
assert is_collection_engine("arxiv") is False
assert is_collection_engine("tavily") is False
class TestClassifyEngineGroup:
def test_collections_win_over_category(self):
# A collection is still "collections" even though it is Local RAG.
assert (
classify_engine_group("collection_1", "Local RAG", False)
== "collections"
)
assert (
classify_engine_group("library", "Local RAG", False)
== "collections"
)
def test_academic(self):
# Free and keyed academic engines both land in academic.
assert classify_engine_group("arxiv", "Scientific", False) == "academic"
assert (
classify_engine_group("nasa_ads", "Scientific", True) == "academic"
)
def test_local_rag_non_collection(self):
assert (
classify_engine_group("elasticsearch", "Local RAG", False)
== "local_rag"
)
def test_books_code_news_have_own_bands_regardless_of_key(self):
assert classify_engine_group("gutenberg", "Books", False) == "books"
# github is Code AND requires a key -> still the Code band, not api_key.
assert classify_engine_group("github", "Code", True) == "code"
assert classify_engine_group("wikinews", "News", False) == "news"
def test_generic_web_split_by_api_key(self):
assert classify_engine_group("ddg", "Web Search", False) == "no_api_key"
assert classify_engine_group("tavily", "Web Search", True) == "api_key"
# The default "Search" category also splits by key.
assert (
classify_engine_group("wikipedia", "Search", False) == "no_api_key"
)
class TestEffectiveGroup:
def test_favorite_overrides_category(self):
assert effective_group("academic", True) == FAVORITES_GROUP_KEY
def test_non_favorite_keeps_base(self):
assert effective_group("academic", False) == "academic"
class TestGroupOrdering:
def test_favorites_sorts_first(self):
assert group_order("favorites") == 0
def test_full_band_order_is_as_specified(self):
expected = [
"favorites",
"collections",
"academic",
"local_rag",
"books",
"code",
"news",
"no_api_key",
"api_key",
]
assert [g.key for g in SEARCH_ENGINE_GROUPS] == expected
# group_order is strictly increasing in that sequence.
orders = [group_order(key) for key in expected]
assert orders == sorted(orders)
assert orders == list(range(len(expected)))
def test_unknown_group_sorts_last(self):
assert group_order("nope") == len(SEARCH_ENGINE_GROUPS)
class TestGroupLabelAndIcon:
def test_known_labels(self):
assert group_label("no_api_key") == "No API key"
assert group_label("api_key") == "API key"
assert group_label("favorites") == "Favorites"
def test_unknown_label_falls_back_to_key(self):
assert group_label("mystery") == "mystery"
def test_icons(self):
assert group_icon("favorites") == "⭐"
assert group_icon("mystery") == ""
class TestEndToEndOrdering:
"""A representative engine set sorts into the intended band order."""
def test_engines_sort_into_bands(self):
# (engine_id, category, requires_api_key, is_favorite)
engines = [
("tavily", "Web Search", True, False), # api_key
("arxiv", "Scientific", False, False), # academic
("collection_7", "Local RAG", False, False), # collections
("ddg", "Web Search", False, False), # no_api_key
("github", "Code", True, False), # code
("elasticsearch", "Local RAG", False, False), # local_rag
("brave", "Web Search", True, True), # favorite (overlay)
("gutenberg", "Books", False, False), # books
("wikinews", "News", False, False), # news
]
options = []
for engine_id, category, key, fav in engines:
base = classify_engine_group(engine_id, category, key)
shown = effective_group(base, fav)
options.append(
{
"value": engine_id,
"label": engine_id,
"group_order": group_order(shown),
}
)
options.sort(key=lambda x: (x["group_order"], x["label"].lower()))
ordered_ids = [o["value"] for o in options]
assert ordered_ids == [
"brave", # favorites
"collection_7", # collections
"arxiv", # academic
"elasticsearch", # local_rag
"gutenberg", # books
"github", # code
"wikinews", # news
"ddg", # no_api_key
"tavily", # api_key
]