Amazon MemoryDB

Vector Search introduction and langchain integration guide.

What is Amazon MemoryDB?

MemoryDB is compatible with Redis OSS, a popular open source data store, enabling you to quickly build applications using the same flexible and friendly Redis OSS data structures, APIs, and commands that they already use today. With MemoryDB, all of your data is stored in memory, which enables you to achieve microsecond read and single-digit millisecond write latency and high throughput. MemoryDB also stores data durably across multiple Availability Zones (AZs) using a Multi-AZ transactional log to enable fast failover, database recovery, and node restarts.

Vector search for MemoryDB

Vector search for MemoryDB extends the functionality of MemoryDB. Vector search can be used in conjunction with existing MemoryDB functionality. Applications that do not use vector search are unaffected by its presence. Vector search is available in all Regions that MemoryDB is available. You can use your existing MemoryDB data or Redis OSS API to build machine learning and generative AI use cases, such as retrieval-augmented generation, anomaly detection, document retrieval, and real-time recommendations.

• Indexing of multiple fields in Redis hashes and JSON
• Vector similarity search (with HNSW (ANN) or FLAT (KNN))
• Vector Range Search (e.g. find all vectors within a radius of a query vector)
• Incremental indexing without performance loss

Setting up

Install Redis Python client

Redis-py is a python client that can be used to connect to MemoryDB

%pip install --upgrade --quiet  redis langchain-aws

from langchain_aws.embeddings import BedrockEmbeddings

embeddings = BedrockEmbeddings()

API Reference:BedrockEmbeddings

MemoryDB Connection

Valid Redis Url schemas are:

1. redis:// - Connection to Redis cluster, unencrypted
2. rediss:// - Connection to Redis cluster, with TLS encryption

More information about additional connection parameters can be found in the redis-py documentation.

Sample data

First we will describe some sample data so that the various attributes of the Redis vector store can be demonstrated.

metadata = [
    {
        "user": "john",
        "age": 18,
        "job": "engineer",
        "credit_score": "high",
    },
    {
        "user": "derrick",
        "age": 45,
        "job": "doctor",
        "credit_score": "low",
    },
    {
        "user": "nancy",
        "age": 94,
        "job": "doctor",
        "credit_score": "high",
    },
    {
        "user": "tyler",
        "age": 100,
        "job": "engineer",
        "credit_score": "high",
    },
    {
        "user": "joe",
        "age": 35,
        "job": "dentist",
        "credit_score": "medium",
    },
]
texts = ["foo", "foo", "foo", "bar", "bar"]
index_name = "users"

Create MemoryDB vector store

The InMemoryVectorStore instance can be initialized using the below methods

• InMemoryVectorStore.__init__ - Initialize directly
• InMemoryVectorStore.from_documents - Initialize from a list of Langchain.docstore.Document objects
• InMemoryVectorStore.from_texts - Initialize from a list of texts (optionally with metadata)
• InMemoryVectorStore.from_existing_index - Initialize from an existing MemoryDB index

from langchain_aws.vectorstores.inmemorydb import InMemoryVectorStore

vds = InMemoryVectorStore.from_texts(
    embeddings,
    redis_url="rediss://cluster_endpoint:6379/ssl=True ssl_cert_reqs=none",
)

API Reference:InMemoryVectorStore

vds.index_name

'users'

Querying

There are multiple ways to query the InMemoryVectorStore implementation based on what use case you have:

• similarity_search: Find the most similar vectors to a given vector.
• similarity_search_with_score: Find the most similar vectors to a given vector and return the vector distance
• similarity_search_limit_score: Find the most similar vectors to a given vector and limit the number of results to the score_threshold
• similarity_search_with_relevance_scores: Find the most similar vectors to a given vector and return the vector similarities
• max_marginal_relevance_search: Find the most similar vectors to a given vector while also optimizing for diversity

results = vds.similarity_search("foo")
print(results[0].page_content)

foo

# with scores (distances)
results = vds.similarity_search_with_score("foo", k=5)
for result in results:
    print(f"Content: {result[0].page_content} --- Score: {result[1]}")

Content: foo --- Score: 0.0
Content: foo --- Score: 0.0
Content: foo --- Score: 0.0
Content: bar --- Score: 0.1566
Content: bar --- Score: 0.1566

# limit the vector distance that can be returned
results = vds.similarity_search_with_score("foo", k=5, distance_threshold=0.1)
for result in results:
    print(f"Content: {result[0].page_content} --- Score: {result[1]}")

Content: foo --- Score: 0.0
Content: foo --- Score: 0.0
Content: foo --- Score: 0.0

# with scores
results = vds.similarity_search_with_relevance_scores("foo", k=5)
for result in results:
    print(f"Content: {result[0].page_content} --- Similiarity: {result[1]}")

Content: foo --- Similiarity: 1.0
Content: foo --- Similiarity: 1.0
Content: foo --- Similiarity: 1.0
Content: bar --- Similiarity: 0.8434
Content: bar --- Similiarity: 0.8434

# you can also add new documents as follows
new_document = ["baz"]
new_metadata = [{"user": "sam", "age": 50, "job": "janitor", "credit_score": "high"}]
# both the document and metadata must be lists
vds.add_texts(new_document, new_metadata)

['doc:users:b9c71d62a0a34241a37950b448dafd38']

MemoryDB as Retriever

Here we go over different options for using the vector store as a retriever.

There are three different search methods we can use to do retrieval. By default, it will use semantic similarity.

query = "foo"
results = vds.similarity_search_with_score(query, k=3, return_metadata=True)

for result in results:
    print("Content:", result[0].page_content, " --- Score: ", result[1])

Content: foo  --- Score:  0.0
Content: foo  --- Score:  0.0
Content: foo  --- Score:  0.0

retriever = vds.as_retriever(search_type="similarity", search_kwargs={"k": 4})

docs = retriever.invoke(query)
docs

[Document(page_content='foo', metadata={'id': 'doc:users_modified:988ecca7574048e396756efc0e79aeca', 'user': 'john', 'job': 'engineer', 'credit_score': 'high', 'age': '18'}),
 Document(page_content='foo', metadata={'id': 'doc:users_modified:009b1afeb4084cc6bdef858c7a99b48e', 'user': 'derrick', 'job': 'doctor', 'credit_score': 'low', 'age': '45'}),
 Document(page_content='foo', metadata={'id': 'doc:users_modified:7087cee9be5b4eca93c30fbdd09a2731', 'user': 'nancy', 'job': 'doctor', 'credit_score': 'high', 'age': '94'}),
 Document(page_content='bar', metadata={'id': 'doc:users_modified:01ef6caac12b42c28ad870aefe574253', 'user': 'tyler', 'job': 'engineer', 'credit_score': 'high', 'age': '100'})]

There is also the similarity_distance_threshold retriever which allows the user to specify the vector distance

retriever = vds.as_retriever(
    search_type="similarity_distance_threshold",
    search_kwargs={"k": 4, "distance_threshold": 0.1},
)

docs = retriever.invoke(query)
docs

[Document(page_content='foo', metadata={'id': 'doc:users_modified:988ecca7574048e396756efc0e79aeca', 'user': 'john', 'job': 'engineer', 'credit_score': 'high', 'age': '18'}),
 Document(page_content='foo', metadata={'id': 'doc:users_modified:009b1afeb4084cc6bdef858c7a99b48e', 'user': 'derrick', 'job': 'doctor', 'credit_score': 'low', 'age': '45'}),
 Document(page_content='foo', metadata={'id': 'doc:users_modified:7087cee9be5b4eca93c30fbdd09a2731', 'user': 'nancy', 'job': 'doctor', 'credit_score': 'high', 'age': '94'})]

Lastly, the similarity_score_threshold allows the user to define the minimum score for similar documents

retriever = vds.as_retriever(
    search_type="similarity_score_threshold",
    search_kwargs={"score_threshold": 0.9, "k": 10},
)

retriever.invoke("foo")

[Document(page_content='foo', metadata={'id': 'doc:users_modified:988ecca7574048e396756efc0e79aeca', 'user': 'john', 'job': 'engineer', 'credit_score': 'high', 'age': '18'}),
 Document(page_content='foo', metadata={'id': 'doc:users_modified:009b1afeb4084cc6bdef858c7a99b48e', 'user': 'derrick', 'job': 'doctor', 'credit_score': 'low', 'age': '45'}),
 Document(page_content='foo', metadata={'id': 'doc:users_modified:7087cee9be5b4eca93c30fbdd09a2731', 'user': 'nancy', 'job': 'doctor', 'credit_score': 'high', 'age': '94'})]

retriever.invoke("foo")

[Document(page_content='foo', metadata={'id': 'doc:users:8f6b673b390647809d510112cde01a27', 'user': 'john', 'job': 'engineer', 'credit_score': 'high', 'age': '18'}),
 Document(page_content='bar', metadata={'id': 'doc:users:93521560735d42328b48c9c6f6418d6a', 'user': 'tyler', 'job': 'engineer', 'credit_score': 'high', 'age': '100'}),
 Document(page_content='foo', metadata={'id': 'doc:users:125ecd39d07845eabf1a699d44134a5b', 'user': 'nancy', 'job': 'doctor', 'credit_score': 'high', 'age': '94'}),
 Document(page_content='foo', metadata={'id': 'doc:users:d6200ab3764c466082fde3eaab972a2a', 'user': 'derrick', 'job': 'doctor', 'credit_score': 'low', 'age': '45'})]

Delete index

To delete your entries you have to address them by their keys.

# delete the indices too
InMemoryVectorStore.drop_index(
    index_name="users", delete_documents=True, redis_url="redis://localhost:6379"
)
InMemoryVectorStore.drop_index(
    index_name="users_modified",
    delete_documents=True,
    redis_url="redis://localhost:6379",
)

True

Related

• Vector store conceptual guide
• Vector store how-to guides