Google Cloud SQL for MySQL
Cloud SQL is a fully managed relational database service that offers high performance, seamless integration, and impressive scalability. It offers PostgreSQL, MySQL, and SQL Server database engines. Extend your database application to build AI-powered experiences leveraging Cloud SQL's LangChain integrations.
This notebook goes over how to useCloud SQL for MySQL
to store vector embeddings with theMySQLVectorStore
class.
Learn more about the package onGitHub.
Before you begin
To run this notebook, you will need to do the following:
- Create a Google Cloud Project
- Enable the Cloud SQL Admin API.
- Create a Cloud SQL instance. (version must be >=8.0.36 withcloudsql_vector database flag configured to "On")
- Create a Cloud SQL database.
- Add a User to the database.
🦜🔗 Library Installation
Install the integration library,langchain-google-cloud-sql-mysql
, and the library for the embedding service,langchain-google-vertexai
.
%pip install--upgrade--quiet langchain-google-cloud-sql-mysql langchain-google-vertexai
Colab only: Uncomment the following cell to restart the kernel or use the button to restart the kernel. For Vertex AI Workbench you can restart the terminal using the button on top.
# # Automatically restart kernel after installs so that your environment can access the new packages
# import IPython
# app = IPython.Application.instance()
# app.kernel.do_shutdown(True)
🔐 Authentication
Authenticate to Google Cloud as the IAM user logged into this notebook in order to access your Google Cloud Project.
- If you are using Colab to run this notebook, use the cell below and continue.
- If you are using Vertex AI Workbench, check out the setup instructionshere.
from google.colabimport auth
auth.authenticate_user()
☁ Set Your Google Cloud Project
Set your Google Cloud project so that you can leverage Google Cloud resources within this notebook.
If you don't know your project ID, try the following:
- Run
gcloud config list
. - Run
gcloud projects list
. - See the support page:Locate the project ID.
# @markdown Please fill in the value below with your Google Cloud project ID and then run the cell.
PROJECT_ID="my-project-id"# @param {type:"string"}
# Set the project id
!gcloud configset project{PROJECT_ID}
Basic Usage
Set Cloud SQL database values
Find your database values, in theCloud SQL Instances page.
Note: MySQL vector support is only available on MySQL instances with version>= 8.0.36.
For existing instances, you may need to perform aself-service maintenance update to update your maintenance version toMYSQL_8_0_36.R20240401.03_00 or greater. Once updated,configure your database flags to have the newcloudsql_vector flag to "On".
# @title Set Your Values Here { display-mode: "form" }
REGION="us-central1"# @param {type: "string"}
INSTANCE="my-mysql-instance"# @param {type: "string"}
DATABASE="my-database"# @param {type: "string"}
TABLE_NAME="vector_store"# @param {type: "string"}
MySQLEngine Connection Pool
One of the requirements and arguments to establish Cloud SQL as a vector store is aMySQLEngine
object. TheMySQLEngine
configures a connection pool to your Cloud SQL database, enabling successful connections from your application and following industry best practices.
To create aMySQLEngine
usingMySQLEngine.from_instance()
you need to provide only 4 things:
project_id
: Project ID of the Google Cloud Project where the Cloud SQL instance is located.region
: Region where the Cloud SQL instance is located.instance
: The name of the Cloud SQL instance.database
: The name of the database to connect to on the Cloud SQL instance.
By default,IAM database authentication will be used as the method of database authentication. This library uses the IAM principal belonging to theApplication Default Credentials (ADC) sourced from the envionment.
For more informatin on IAM database authentication please see:
Optionally,built-in database authentication using a username and password to access the Cloud SQL database can also be used. Just provide the optionaluser
andpassword
arguments toMySQLEngine.from_instance()
:
user
: Database user to use for built-in database authentication and loginpassword
: Database password to use for built-in database authentication and login.
from langchain_google_cloud_sql_mysqlimport MySQLEngine
engine= MySQLEngine.from_instance(
project_id=PROJECT_ID, region=REGION, instance=INSTANCE, database=DATABASE
)
Initialize a table
TheMySQLVectorStore
class requires a database table. TheMySQLEngine
class has a helper methodinit_vectorstore_table()
that can be used to create a table with the proper schema for you.
engine.init_vectorstore_table(
table_name=TABLE_NAME,
vector_size=768,# Vector size for VertexAI model(textembedding-gecko@latest)
)
Create an embedding class instance
You can use anyLangChain embeddings model.You may need to enable the Vertex AI API to useVertexAIEmbeddings
.
We recommend pinning the embedding model's version for production, learn more about theText embeddings models.
# enable Vertex AI API
!gcloud services enable aiplatform.googleapis.com
from langchain_google_vertexaiimport VertexAIEmbeddings
embedding= VertexAIEmbeddings(
model_name="textembedding-gecko@latest", project=PROJECT_ID
)
Initialize a default MySQLVectorStore
To initialize aMySQLVectorStore
class you need to provide only 3 things:
engine
- An instance of aMySQLEngine
engine.embedding_service
- An instance of a LangChain embedding model.table_name
: The name of the table within the Cloud SQL database to use as the vector store.
from langchain_google_cloud_sql_mysqlimport MySQLVectorStore
store= MySQLVectorStore(
engine=engine,
embedding_service=embedding,
table_name=TABLE_NAME,
)
Add texts
import uuid
all_texts=["Apples and oranges","Cars and airplanes","Pineapple","Train","Banana"]
metadatas=[{"len":len(t)}for tin all_texts]
ids=[str(uuid.uuid4())for _in all_texts]
store.add_texts(all_texts, metadatas=metadatas, ids=ids)
Delete texts
Delete vectors from the vector store by ID.
store.delete([ids[1]])
Search for documents
query="I'd like a fruit."
docs= store.similarity_search(query)
print(docs[0].page_content)
Pineapple
Search for documents by vector
It is also possible to do a search for documents similar to a given embedding vector usingsimilarity_search_by_vector
which accepts an embedding vector as a parameter instead of a string.
query_vector= embedding.embed_query(query)
docs= store.similarity_search_by_vector(query_vector, k=2)
print(docs)
[Document(page_content='Pineapple', metadata={'len': 9}), Document(page_content='Banana', metadata={'len': 6})]
Add an index
Speed up vector search queries by applying a vector index. Learn more aboutMySQL vector indexes.
Note: For IAM database authentication (default usage), the IAM database user will need to be granted the following permissions by a privileged database user for full control of vector indexes.
GRANT EXECUTE ON PROCEDURE mysql.create_vector_index TO '<IAM_DB_USER>'@'%';
GRANT EXECUTE ON PROCEDURE mysql.alter_vector_index TO '<IAM_DB_USER>'@'%';
GRANT EXECUTE ON PROCEDURE mysql.drop_vector_index TO '<IAM_DB_USER>'@'%';
GRANT SELECT ON mysql.vector_indexes TO '<IAM_DB_USER>'@'%';
from langchain_google_cloud_sql_mysqlimport VectorIndex
store.apply_vector_index(VectorIndex())
Remove an index
store.drop_vector_index()
Advanced Usage
Create a MySQLVectorStore with custom metadata
A vector store can take advantage of relational data to filter similarity searches.
Create a table andMySQLVectorStore
instance with custom metadata columns.
from langchain_google_cloud_sql_mysqlimport Column
# set table name
CUSTOM_TABLE_NAME="vector_store_custom"
engine.init_vectorstore_table(
table_name=CUSTOM_TABLE_NAME,
vector_size=768,# VertexAI model: textembedding-gecko@latest
metadata_columns=[Column("len","INTEGER")],
)
# initialize MySQLVectorStore with custom metadata columns
custom_store= MySQLVectorStore(
engine=engine,
embedding_service=embedding,
table_name=CUSTOM_TABLE_NAME,
metadata_columns=["len"],
# connect to an existing VectorStore by customizing the table schema:
# id_column="uuid",
# content_column="documents",
# embedding_column="vectors",
)
Search for documents with metadata filter
It can be helpful to narrow down the documents before working with them.
For example, documents can be filtered on metadata using thefilter
argument.
import uuid
# add texts to the vector store
all_texts=["Apples and oranges","Cars and airplanes","Pineapple","Train","Banana"]
metadatas=[{"len":len(t)}for tin all_texts]
ids=[str(uuid.uuid4())for _in all_texts]
custom_store.add_texts(all_texts, metadatas=metadatas, ids=ids)
# use filter on search
query_vector= embedding.embed_query("I'd like a fruit.")
docs= custom_store.similarity_search_by_vector(query_vector,filter="len >= 6")
print(docs)
[Document(page_content='Pineapple', metadata={'len': 9}), Document(page_content='Banana', metadata={'len': 6}), Document(page_content='Apples and oranges', metadata={'len': 18}), Document(page_content='Cars and airplanes', metadata={'len': 18})]
Related
- Vector storeconceptual guide
- Vector storehow-to guides