Skip to main content
This notebook covers how to get started with the Oceanbase vector store.

​
Setup

To access Oceanbase vector stores you’ll need to deploy a standalone OceanBase server: %docker run —name=ob433 -e MODE=mini -e OB_SERVER_IP=127.0.0.1 -p 2881:2881 -d quay.io/oceanbase/oceanbase-ce:4.3.3.1-101000012024102216 And install the langchain-oceanbase integration package. pip install -qU “langchain-oceanbase” Check the connection to OceanBase and set the memory usage ratio for vector data: 
from pyobvector import ObVecClient

tmp_client = ObVecClient()
tmp_client.perform_raw_text_sql("ALTER SYSTEM ob_vector_memory_limit_percentage = 30")

<sqlalchemy.engine.cursor.CursorResult at 0x12696f2a0>

​
Initialization

Configure the API key of the embedded model. Here we use DashScopeEmbeddings as an example. When deploying Oceanbase with a Docker image as described above, simply follow the script below to set the host, port, user, password, and database name. For other deployment methods, set these parameters according to the actual situation. pip install dashscope 
import os

from langchain_community.embeddings import DashScopeEmbeddings
from langchain_oceanbase.vectorstores import OceanbaseVectorStore

DASHSCOPE_API = os.environ.get("DASHSCOPE_API_KEY", "")
connection_args = {
    "host": "127.0.0.1",
    "port": "2881",
    "user": "root@test",
    "password": "",
    "db_name": "test",
}

embeddings = DashScopeEmbeddings(
    model="text-embedding-v1", dashscope_api_key=DASHSCOPE_API
)

vector_store = OceanbaseVectorStore(
    embedding_function=embeddings,
    table_name="langchain_vector",
    connection_args=connection_args,
    vidx_metric_type="l2",
    drop_old=True,
)

​
Manage vector store

​
Add items to vector store

  • TODO: Edit and then run code cell to generate output
from langchain_core.documents import Document

document_1 = Document(page_content="foo", metadata={"source": "https://foo.com"})
document_2 = Document(page_content="bar", metadata={"source": "https://bar.com"})
document_3 = Document(page_content="baz", metadata={"source": "https://baz.com"})

documents = [document_1, document_2, document_3]

vector_store.add_documents(documents=documents, ids=["1", "2", "3"])

['1', '2', '3']

​
Update items in vector store

updated_document = Document(
    page_content="qux", metadata={"source": "https://another-example.com"}
)

vector_store.add_documents(documents=[updated_document], ids=["1"])

['1']

​
Delete items from vector store

vector_store.delete(ids=["3"])

​
Query vector store

Once your vector store has been created and the relevant documents have been added you will most likely wish to query it during the running of your chain or agent.

​
Query directly

Performing a simple similarity search can be done as follows: 
results = vector_store.similarity_search(
    query="thud", k=1, filter={"source": "https://another-example.com"}
)
for doc in results:
    print(f"* {doc.page_content} [{doc.metadata}]")

* bar [{'source': 'https://bar.com'}]
If you want to execute a similarity search and receive the corresponding scores you can run: 
results = vector_store.similarity_search_with_score(
    query="thud", k=1, filter={"source": "https://example.com"}
)
for doc, score in results:
    print(f"* [SIM={score:3f}] {doc.page_content} [{doc.metadata}]")

* [SIM=133.452299] bar [{'source': 'https://bar.com'}]

​
Query by turning into retriever

You can also transform the vector store into a retriever for easier usage in your chains. 
retriever = vector_store.as_retriever(search_kwargs={"k": 1})
retriever.invoke("thud")

[Document(metadata={'source': 'https://bar.com'}, page_content='bar')]

​
Usage for retrieval-augmented generation

For guides on how to use this vector store for retrieval-augmented generation (RAG), see the following sections:

​
API reference

For detailed documentation of all OceanbaseVectorStore features and configurations head to the API reference: python.langchain.com/docs/integrations/vectorstores/oceanbase
Edit the source of this page on GitHub.
Connect these docs programmatically to Claude, VSCode, and more via MCP for real-time answers.