55 lines
1.9 KiB
Python
55 lines
1.9 KiB
Python
import os
|
|
|
|
import pytest
|
|
|
|
from guardrails.document_store import EphemeralDocumentStore
|
|
from guardrails.embedding import OpenAIEmbedding
|
|
from guardrails.vectordb import Faiss
|
|
|
|
|
|
@pytest.mark.skipif(
|
|
os.environ.get("OPENAI_API_KEY") in [None, "mocked"],
|
|
reason="openai api key not set",
|
|
)
|
|
class TestEphemeralDocumentStore:
|
|
def test_similarity_search(self):
|
|
sentences = {
|
|
"who is the best player in the world": "Kevin Durant",
|
|
"who is the current mvp": "Steph Curry",
|
|
"who is on mount rushmore of nba": "Lebron James, Michael Jordan",
|
|
}
|
|
db = Faiss.new_flat_l2_index(1536, OpenAIEmbedding())
|
|
store = EphemeralDocumentStore(db)
|
|
_ = [
|
|
store.add_text(text, {"ctx": addn_ctx})
|
|
for text, addn_ctx in sentences.items()
|
|
]
|
|
pages = store.search("mvp", 1)
|
|
assert len(pages) == 1
|
|
assert pages[0].text == "who is the current mvp"
|
|
assert pages[0].metadata["ctx"] == "Steph Curry"
|
|
|
|
def test_batched_add(self):
|
|
db = Faiss.new_flat_l2_index(1536, OpenAIEmbedding())
|
|
store = EphemeralDocumentStore(db)
|
|
new_doc_ids = store.add_texts({"foo": {"ctx": "bar"}, "pipe": {"ctx": "baz"}})
|
|
assert len(new_doc_ids) == 2
|
|
|
|
def test_persistence(self):
|
|
if os.path.exists("test.db"):
|
|
os.remove("test.db")
|
|
if os.path.exists("test.index"):
|
|
os.remove("test.index")
|
|
|
|
db = Faiss.new_flat_l2_index(1536, OpenAIEmbedding(), "test.index")
|
|
store = EphemeralDocumentStore(db, "test.db")
|
|
doc_id = store.add_text("foo", {"ctx": "bar"})
|
|
store.add_text("foo", {"ctx": "bar"})
|
|
assert doc_id is not None
|
|
store.flush()
|
|
|
|
db2 = Faiss.load("test.index", OpenAIEmbedding())
|
|
store2 = EphemeralDocumentStore(db2, "test.db")
|
|
pages = store2.search("foo")
|
|
assert len(pages) == 1
|