parameterized distance metrics; lint; format; tests (#4375)

# Parameterize Redis vectorstore index

Redis vectorstore allows for three different distance metrics: `L2`
(flat L2), `COSINE`, and `IP` (inner product). Currently, the
`Redis._create_index` method hard codes the distance metric to COSINE.

I've parameterized this as an argument in the `Redis.from_texts` method
-- pretty simple.

Fixes #4368 

## Before submitting

I've added an integration test showing indexes can be instantiated with
all three values in the `REDIS_DISTANCE_METRICS` literal. An example
notebook seemed overkill here. Normal API documentation would be more
appropriate, but no standards are in place for that yet.

## Who can review?

Not sure who's responsible for the vectorstore module... Maybe @eyurtsev
/ @hwchase17 / @agola11 ?
This commit is contained in:
Evan Jones 2023-05-11 03:20:01 -04:00 committed by GitHub
parent f46710d408
commit f668251948
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 54 additions and 6 deletions

View File

@ -11,6 +11,7 @@ from typing import (
Dict,
Iterable,
List,
Literal,
Mapping,
Optional,
Tuple,
@ -39,6 +40,9 @@ REDIS_REQUIRED_MODULES = [
{"name": "searchlight", "ver": 20400},
]
# distance mmetrics
REDIS_DISTANCE_METRICS = Literal["COSINE", "IP", "L2"]
def _check_redis_module_exist(client: RedisType, required_modules: List[dict]) -> None:
"""Check if the correct Redis modules are installed."""
@ -142,7 +146,9 @@ class Redis(VectorStore):
self.vector_key = vector_key
self.relevance_score_fn = relevance_score_fn
def _create_index(self, dim: int = 1536) -> None:
def _create_index(
self, dim: int = 1536, distance_metric: REDIS_DISTANCE_METRICS = "COSINE"
) -> None:
try:
from redis.commands.search.field import TextField, VectorField
from redis.commands.search.indexDefinition import IndexDefinition, IndexType
@ -154,10 +160,7 @@ class Redis(VectorStore):
# Check if index exists
if not _check_index_exists(self.client, self.index_name):
# Constants
distance_metric = (
"COSINE" # distance metric for the vectors (ex. COSINE, IP, L2)
)
# Define schema
schema = (
TextField(name=self.content_key),
TextField(name=self.metadata_key),
@ -364,6 +367,7 @@ class Redis(VectorStore):
content_key: str = "content",
metadata_key: str = "metadata",
vector_key: str = "content_vector",
distance_metric: REDIS_DISTANCE_METRICS = "COSINE",
**kwargs: Any,
) -> Redis:
"""Create a Redis vectorstore from raw documents.
@ -407,7 +411,7 @@ class Redis(VectorStore):
embeddings = embedding.embed_documents(texts)
# Create the search index
instance._create_index(dim=len(embeddings[0]))
instance._create_index(dim=len(embeddings[0]), distance_metric=distance_metric)
# Add data to Redis
instance.add_texts(texts, metadatas, embeddings)

View File

@ -1,4 +1,6 @@
"""Test Redis functionality."""
import pytest
from langchain.docstore.document import Document
from langchain.vectorstores.redis import Redis
from tests.integration_tests.vectorstores.fake_embeddings import FakeEmbeddings
@ -7,6 +9,9 @@ TEST_INDEX_NAME = "test"
TEST_REDIS_URL = "redis://localhost:6379"
TEST_SINGLE_RESULT = [Document(page_content="foo")]
TEST_RESULT = [Document(page_content="foo"), Document(page_content="foo")]
COSINE_SCORE = pytest.approx(0.05, abs=0.002)
IP_SCORE = -8.0
EUCLIDEAN_SCORE = 1.0
def drop(index_name: str) -> bool:
@ -58,3 +63,42 @@ def test_redis_add_texts_to_existing() -> None:
output = docsearch.similarity_search("foo", k=2)
assert output == TEST_RESULT
assert drop(TEST_INDEX_NAME)
def test_cosine() -> None:
"""Test cosine distance."""
texts = ["foo", "bar", "baz"]
docsearch = Redis.from_texts(
texts,
FakeEmbeddings(),
redis_url=TEST_REDIS_URL,
distance_metric="COSINE",
)
output = docsearch.similarity_search_with_score("far", k=2)
_, score = output[1]
assert score == COSINE_SCORE
assert drop(docsearch.index_name)
def test_l2() -> None:
"""Test Flat L2 distance."""
texts = ["foo", "bar", "baz"]
docsearch = Redis.from_texts(
texts, FakeEmbeddings(), redis_url=TEST_REDIS_URL, distance_metric="L2"
)
output = docsearch.similarity_search_with_score("far", k=2)
_, score = output[1]
assert score == EUCLIDEAN_SCORE
assert drop(docsearch.index_name)
def test_ip() -> None:
"""Test inner product distance."""
texts = ["foo", "bar", "baz"]
docsearch = Redis.from_texts(
texts, FakeEmbeddings(), redis_url=TEST_REDIS_URL, distance_metric="IP"
)
output = docsearch.similarity_search_with_score("far", k=2)
_, score = output[1]
assert score == IP_SCORE
assert drop(docsearch.index_name)