Embedding

Embeddings serve as the A.I.-native approach for representing various data types, including text, images, and soon, audio and video, making them the perfect fit for integration with A.I.-powered tools and algorithms.

Within the Index Network, an Item can have many embeddings which serves as multi-dimensional representations for items in specific contexts.

Storing embeddings in a decentralized way efficiently promotes a cohesive and interconnected knowledge network for A.I.-driven applications.

Schema

The Embedding schema consists of several key properties:

Property
Description
Type

modelName

Name of the model used for generating the embedding.

String

vector

The embedding vector, represented as an array of floats.

Array of Floats

context

Context information associated with the embedding, detailed in the EmbeddingContext model.

EmbeddingContext

indexId

Identifier of the Index associated with this embedding.

StreamID

index

The actual Index object associated with this embedding, accessed through indexId.

Index

itemId

Identifier of the item (node) associated with this embedding.

StreamID

item

The actual Node object associated with this embedding, accessed through itemId.

Node

createdAt

The timestamp when the Embedding was initially created, for historical record-keeping.

DateTime

updatedAt

The timestamp when the Embedding was last updated, ensuring data accuracy.

DateTime

deletedAt

The timestamp when the Embedding was deleted (if applicable).

DateTime

controllerDID

The Decentralized Identifier (DID) of the controller for access control.

DID

version

The CommitID representing the version or state of the Embedding for version control.

CommitID

Context Schema

The EmbeddingContext type in a data model plays a crucial role in enhancing the interoperability of embeddings. It provides a standardized and enriched representation of data, such as a node's content augmented with additional context, ensuring that embeddings are universally applicable across different systems. T

Property
Description

context

The transformed representation of a node before obtaining embeddings. This includes the node's content with added context information for embedding tasks like summaries, etc. Raw document is used if it's null.

description

A human-readable description of the related context. It helps in understanding the nature and purpose of the context used for embedding.

category

The category or namespace for the embedding. Example values might include "summaries", "knowledge_graph", "document", etc., indicating the domain or type of embedding being used.

Endpoints

This section provides details about the available API endpoints.

Last updated