Embeddings

Super Admin

What are Embeddings?

Embeddings are vector representations of documents imported into the knowledge base. They allow AI agents to perform semantic searches — finding relevant information based on the meaning of the content, not just keywords.

Each processed document generates one or more embeddings that are stored in the workspace’s vector database.

Accessing the Tab

The Embeddings tab is displayed in the workspace’s Knowledge Base section, alongside the Overview, Storage, and Pipelines tabs. It is only visible to users with a system profile.

Embeddings Table

The table displays all generated embeddings with the following columns:

Column	Description
ID	Unique embedding identifier (truncated UUID)
Inserted At	Embedding creation date and time
Metadata	Associated key-value pairs displayed as badges
Version	Embedding version number
Actions	View details and delete

Pagination

Displays 10 items per page by default
Configurable sizes: 5, 10, 20, 50, or 100 items per page
Shows “Showing X to Y of Z embeddings” summary
Page navigation controls

Search and Filters

Text search — search by embedding ID
Date search — filter by insertion date using a date picker
Column visibility — show or hide table columns (ID, Metadata, Version, Inserted At)

Viewing Details

Click the view icon to open a side panel with the complete embedding information:

Basic information — ID, Version, Inserted At, Knowledge Base ID
Content — full embedding text in a scrollable area
Metadata — complete JSON metadata in formatted view

Deleting Embeddings

Click the delete icon to remove an individual embedding.

A confirmation dialog is displayed before deletion
The action is irreversible — the embedding will be permanently removed from the vector database
The table refreshes automatically after deletion