Managed Tensor Database

Deep Lake Managed Database

Overview of Deep Lake's Managed Tensor Database

Deep Lake offers a serverless Managed Tensor Database to eliminate the complexities of self-hosting and substantially lowers costs. Currently, it only supports dataset queries, including vector search, but additional features for creating and modifying data being added.

User Interfaces

LangChain and LlamaIndex

No changes are needed from the user to use the Managed Tensor Database in LangChain or LlamaIndex, since all of the methods automatically route request to the Managed Database when needed.

The only requirement it to specify that the Vector Store should be stored in the Managed Database by specifying dataset_path = hub://org_id/dataset_name and runtime = {"tensor_db": True} during Vector Store creation.

REST API

A standalone REST API is available for interacting with the Managed Database:

REST API

Architecture

The Managed Tensor Database is serverless and can deployed in the user's VPC.

DETAILS COMING SOON

Further Information:

Migrating Datasets to the Tensor Database