mirror of
https://github.com/hwchase17/langchain
synced 2024-10-29 17:07:25 +00:00
112 lines
4.1 KiB
Python
112 lines
4.1 KiB
Python
"""Test FAISS functionality."""
|
|
import tempfile
|
|
|
|
import pytest
|
|
|
|
from langchain.docstore.document import Document
|
|
from langchain.docstore.in_memory import InMemoryDocstore
|
|
from langchain.docstore.wikipedia import Wikipedia
|
|
from langchain.vectorstores.faiss import FAISS
|
|
from tests.integration_tests.vectorstores.fake_embeddings import FakeEmbeddings
|
|
|
|
|
|
def test_faiss() -> None:
|
|
"""Test end to end construction and search."""
|
|
texts = ["foo", "bar", "baz"]
|
|
docsearch = FAISS.from_texts(texts, FakeEmbeddings())
|
|
index_to_id = docsearch.index_to_docstore_id
|
|
expected_docstore = InMemoryDocstore(
|
|
{
|
|
index_to_id[0]: Document(page_content="foo"),
|
|
index_to_id[1]: Document(page_content="bar"),
|
|
index_to_id[2]: Document(page_content="baz"),
|
|
}
|
|
)
|
|
assert docsearch.docstore.__dict__ == expected_docstore.__dict__
|
|
output = docsearch.similarity_search("foo", k=1)
|
|
assert output == [Document(page_content="foo")]
|
|
|
|
|
|
def test_faiss_vector_sim() -> None:
|
|
"""Test vector similarity."""
|
|
texts = ["foo", "bar", "baz"]
|
|
docsearch = FAISS.from_texts(texts, FakeEmbeddings())
|
|
index_to_id = docsearch.index_to_docstore_id
|
|
expected_docstore = InMemoryDocstore(
|
|
{
|
|
index_to_id[0]: Document(page_content="foo"),
|
|
index_to_id[1]: Document(page_content="bar"),
|
|
index_to_id[2]: Document(page_content="baz"),
|
|
}
|
|
)
|
|
assert docsearch.docstore.__dict__ == expected_docstore.__dict__
|
|
query_vec = FakeEmbeddings().embed_query(text="foo")
|
|
output = docsearch.similarity_search_by_vector(query_vec, k=1)
|
|
assert output == [Document(page_content="foo")]
|
|
|
|
# make sure we can have k > docstore size
|
|
output = docsearch.max_marginal_relevance_search_by_vector(query_vec, k=10)
|
|
assert len(output) == len(texts)
|
|
|
|
|
|
def test_faiss_with_metadatas() -> None:
|
|
"""Test end to end construction and search."""
|
|
texts = ["foo", "bar", "baz"]
|
|
metadatas = [{"page": i} for i in range(len(texts))]
|
|
docsearch = FAISS.from_texts(texts, FakeEmbeddings(), metadatas=metadatas)
|
|
expected_docstore = InMemoryDocstore(
|
|
{
|
|
docsearch.index_to_docstore_id[0]: Document(
|
|
page_content="foo", metadata={"page": 0}
|
|
),
|
|
docsearch.index_to_docstore_id[1]: Document(
|
|
page_content="bar", metadata={"page": 1}
|
|
),
|
|
docsearch.index_to_docstore_id[2]: Document(
|
|
page_content="baz", metadata={"page": 2}
|
|
),
|
|
}
|
|
)
|
|
assert docsearch.docstore.__dict__ == expected_docstore.__dict__
|
|
output = docsearch.similarity_search("foo", k=1)
|
|
assert output == [Document(page_content="foo", metadata={"page": 0})]
|
|
|
|
|
|
def test_faiss_search_not_found() -> None:
|
|
"""Test what happens when document is not found."""
|
|
texts = ["foo", "bar", "baz"]
|
|
docsearch = FAISS.from_texts(texts, FakeEmbeddings())
|
|
# Get rid of the docstore to purposefully induce errors.
|
|
docsearch.docstore = InMemoryDocstore({})
|
|
with pytest.raises(ValueError):
|
|
docsearch.similarity_search("foo")
|
|
|
|
|
|
def test_faiss_add_texts() -> None:
|
|
"""Test end to end adding of texts."""
|
|
# Create initial doc store.
|
|
texts = ["foo", "bar", "baz"]
|
|
docsearch = FAISS.from_texts(texts, FakeEmbeddings())
|
|
# Test adding a similar document as before.
|
|
docsearch.add_texts(["foo"])
|
|
output = docsearch.similarity_search("foo", k=2)
|
|
assert output == [Document(page_content="foo"), Document(page_content="foo")]
|
|
|
|
|
|
def test_faiss_add_texts_not_supported() -> None:
|
|
"""Test adding of texts to a docstore that doesn't support it."""
|
|
docsearch = FAISS(FakeEmbeddings().embed_query, None, Wikipedia(), {})
|
|
with pytest.raises(ValueError):
|
|
docsearch.add_texts(["foo"])
|
|
|
|
|
|
def test_faiss_local_save_load() -> None:
|
|
"""Test end to end serialization."""
|
|
texts = ["foo", "bar", "baz"]
|
|
docsearch = FAISS.from_texts(texts, FakeEmbeddings())
|
|
|
|
with tempfile.NamedTemporaryFile() as temp_file:
|
|
docsearch.save_local(temp_file.name)
|
|
new_docsearch = FAISS.load_local(temp_file.name, FakeEmbeddings())
|
|
assert new_docsearch.index is not None
|