Skip to main content
Tablestore is a fully managed NoSQL cloud database service. Tablestore enables storage of a massive amount of structured and semi-structured data. This notebook shows how to use functionality related to the Tablestore vector database. To use Tablestore, you must create an instance. Here are the creating instance instructions.

​
Setup

%pip install -qU langchain-community tablestore

​
Initialization

import getpass import os  os.environ["end_point"] = getpass.getpass("Tablestore end_point:") os.environ["instance_name"] = getpass.getpass("Tablestore instance_name:") os.environ["access_key_id"] = getpass.getpass("Tablestore access_key_id:") os.environ["access_key_secret"] = getpass.getpass("Tablestore access_key_secret:")
Create vector store. 
import tablestore from langchain_community.embeddings import FakeEmbeddings from langchain_community.vectorstores import TablestoreVectorStore from langchain_core.documents import Document  test_embedding_dimension_size = 4 embeddings = FakeEmbeddings(size=test_embedding_dimension_size)  store = TablestoreVectorStore(  embedding=embeddings,  endpoint=os.getenv("end_point"),  instance_name=os.getenv("instance_name"),  access_key_id=os.getenv("access_key_id"),  access_key_secret=os.getenv("access_key_secret"),  vector_dimension=test_embedding_dimension_size,  # metadata mapping is used to filter non-vector fields.  metadata_mappings=[  tablestore.FieldSchema(  "type", tablestore.FieldType.KEYWORD, index=True, enable_sort_and_agg=True  ),  tablestore.FieldSchema(  "time", tablestore.FieldType.LONG, index=True, enable_sort_and_agg=True  ),  ], )

​
Manage vector store

Create table and index. 
store.create_table_if_not_exist() store.create_search_index_if_not_exist()
Add documents. 
store.add_documents(  [  Document(  id="1", page_content="1 hello world", metadata={"type": "pc", "time": 2000}  ),  Document(  id="2", page_content="abc world", metadata={"type": "pc", "time": 2009}  ),  Document(  id="3", page_content="3 text world", metadata={"type": "sky", "time": 2010}  ),  Document(  id="4", page_content="hi world", metadata={"type": "sky", "time": 2030}  ),  Document(  id="5", page_content="hi world", metadata={"type": "sky", "time": 2030}  ),  ] ) 
['1', '2', '3', '4', '5']
Delete document. 
store.delete(["3"]) 
True
Get documents.

​
Query vector store

store.get_by_ids(["1", "3", "5"]) 
[Document(id='1', metadata={'embedding': '[1.3296732307905934, 0.0037521341868022385, 0.9821875819319514, 2.5644103644492393]', 'time': 2000, 'type': 'pc'}, page_content='1 hello world'),  None,  Document(id='5', metadata={'embedding': '[1.4558082172139821, -1.6441137122167426, -0.13113098640337423, -1.889685473174525]', 'time': 2030, 'type': 'sky'}, page_content='hi world')]
Similarity search. 
store.similarity_search(query="hello world", k=2) 
[Document(id='1', metadata={'embedding': [1.3296732307905934, 0.0037521341868022385, 0.9821875819319514, 2.5644103644492393], 'time': 2000, 'type': 'pc'}, page_content='1 hello world'),  Document(id='4', metadata={'embedding': [-0.3310144199800685, 0.29250046478723635, -0.0646862290377582, -0.23664360156781225], 'time': 2030, 'type': 'sky'}, page_content='hi world')]
Similarity search with filters. 
store.similarity_search(  query="hello world",  k=10,  tablestore_filter_query=tablestore.BoolQuery(  must_queries=[tablestore.TermQuery(field_name="type", column_value="sky")],  should_queries=[tablestore.RangeQuery(field_name="time", range_from=2020)],  must_not_queries=[tablestore.TermQuery(field_name="type", column_value="pc")],  ), ) 
[Document(id='5', metadata={'embedding': [1.4558082172139821, -1.6441137122167426, -0.13113098640337423, -1.889685473174525], 'time': 2030, 'type': 'sky'}, page_content='hi world'),  Document(id='4', metadata={'embedding': [-0.3310144199800685, 0.29250046478723635, -0.0646862290377582, -0.23664360156781225], 'time': 2030, 'type': 'sky'}, page_content='hi world')]

​
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 TablestoreVectorStore features and configurations head to the API reference: python.langchain.com/api_reference/community/vectorstores/langchain_community.vectorstores.tablestore.TablestoreVectorStore.html
Edit the source of this page on GitHub
⌘I