Skip to main content
The Data Layer is AstroBee’s semantic modeling feature. It creates a business-friendly abstraction over your raw data, allowing you to ask questions in plain English without knowing where the data lives or how tables connect.

Overview

A data layer consists of:
  • Entities — Business concepts like Customer, Product, or Order
  • Properties — Attributes of entities (e.g., Customer.email, Product.price)
  • Relationships — How entities connect (e.g., Customer → Order)
Data layer graph

Accessing the Data Layer

Click Data Layer in the sidebar to open the editor. The editor has three main tabs:
TabPurpose
Data LayerView and edit entities in Graph or List view
SourcesConnect data sources to your data layer
LinkagesView discovered relationships between tables

Generating a Data Layer

Connect Sources First

Before generating, connect at least one data source:
  1. Click the Sources tab
  2. Find available data sources
  3. Click Connect to link them to this data layer
Sources tab

AI-Assisted Generation

Click Generate Data Layer from selected Sources to start the AI-assisted workflow:
  1. Business Questions — AstroBee asks what you want to analyze (customer behavior, sales performance, etc.)
  2. Terminology — Define your business language (orders vs. sales vs. transactions)
  3. Generation — AI analyzes your data and creates entities with proper bindings
Generation questionnaire
The AI will:
  • Discover tables and columns in your data sources
  • Match entities to tables by name similarity
  • Map properties to columns based on data types and meanings
  • Create relationships based on discovered linkages
Generation summary

Viewing Your Data Layer

Graph View

The default view shows entities as nodes and relationships as edges:
Graph view
  • Nodes represent entities (dimension, fact, or bridge types)
  • Edges show relationships with cardinality (1:N, N:N)
  • Click Expand to fullscreen for a larger view

List View

Click List View for a detailed breakdown:
List view
The list view shows:
  • Entity names and descriptions
  • All properties with types (Dimension/Measure) and data types
  • Field descriptions
  • Relationships with join conditions

Editing Your Data Layer

Chat-Based Editing

Use the chat input at the bottom to make changes in natural language:
Add a property called "lifetime_value" to Customer that calculates
total spend across all orders
Create a relationship between Product and OrderItem
AstroBee interprets your request, makes the changes, and shows them in the Review panel.

Direct Editing

In List View, click on any element to edit directly:
  • Click entity names to rename
  • Click descriptions to update
  • Click properties to modify types or descriptions

Reviewing Changes

All changes go to a scratchpad before being published. Click the Review button to see pending changes:
Review panel
The review panel shows:
  • Created items (new entities, properties, relationships)
  • Modified items (updated definitions)
  • Deleted items (removed elements)
Actions:
  • Publish all pending changes — Apply changes and create a new version
  • Discard all pending changes — Revert the scratchpad

Version History

Every time you publish, AstroBee creates a new version. View history by clicking the version buttons:
Version history
Version history includes:
  • Timestamp of each version
  • Conversation history (what was discussed to create that version)
  • Ability to view any previous version
Versions are immutable. To make changes, edit the current version and publish a new one.

Linkage Discovery

The Linkages tab shows relationships AstroBee discovered between your source tables:
Linkages tab
Click a linkage to see details:
Linkage details
Each linkage shows:
  • SQL join logic — The actual join condition
  • Validation status — Whether the join was tested
  • Statistics — Confidence, coverage, row counts
  • Sample matches — Example values that match

Entity Types

AstroBee uses three entity types:
TypePurposeExample
DimensionDescriptive attributes for grouping/filteringCustomer, Product, Region
FactMeasurable events or transactionsOrder, PageView, Payment
BridgeJunction tables for many-to-many relationshipsOrderItem, UserRole

Property Types

Properties are classified as:
TypePurposeExample
DimensionAttributes for grouping or filteringcustomer_id, category, status
MeasureNumeric values for aggregationprice, quantity, total_amount

Best Practices

Let AstroBee create the initial structure, then refine. It’s faster than building from scratch.
Entity and property names should match how your team talks about the business.
Good descriptions help AstroBee understand context when answering questions.
Check the Linkages tab to ensure discovered relationships are correct before building on them.
Publish often to maintain a clear history of changes.

Next Steps