Acceldata
ADOC

Last updated: Nov 13, 2026 04:27 UTC

Glossary

Agent: An AI component that handles a specific type of data management task, for example, the Data Quality Agent or the Catalog Agent.

Asset: A data object such as a table, database, file, or dataset managed within your data catalog.

Conversation: An interactive session between a user and ADM where context is maintained across multiple messages.

Knowledge Base: A shared library of your organization's documents, such as policies and runbooks. ADM searches these documents to answer questions with your organization's own content. Every user in your tenant can search Knowledge Base files.

LLM (Large Language Model): The AI model that powers ADM's natural language understanding and generation capabilities.

MCP (Model Context Protocol): Open standard for connecting ADM with external tools and services like Google Drive and Gmail.

Notebook: Collection of queries that can be executed together and scheduled for regular runs.

Policy: Set of data quality, governance, or compliance rules applied to data assets.

Policy Score: Percentage indicating overall data quality health, calculated from policy execution results.

Prompt: Text input provided to ADM to elicit a response or action.

Response Style: A setting that controls how ADM presents its answers. The available styles are Executive Summary, Balanced Insight (the default), Raw Data, and Technical Deep Dive.

Slash Command: A command that starts with /, such as /Visualize Pipeline. A slash command runs a specific action or workflow.

Thinking: The steps that ADM shows while it works on your question, such as the agents and tools that it uses. To see how ADM reached an answer, expand Thinking under the answer.

Token: Text chunk processed by LLM; approximately 4 characters equal 1 token.

Workflow: Multi-step automated process that guides users through complex data management tasks.

Workspace Memory: Key findings that users save from conversations with Remember for this workspace. ADM uses these findings in later conversations about the same entity or topic.