Skip to main content
The Chat page in Edit Data Layer mode shows how AstroBee understands your business data model. Here you can explore entities, view their properties and relationships, and make changes using natural language to improve how AstroBee interprets your data.

Interface overview

This page has three main sections:
Data Layer interface showing entity list and properties
  • Chat - The current active section (highlighted ) where you can edit your data layer
  • Sources - View raw data connections
  • Conversations - See your past questions and AstroBee’s responses

Left entity browser

  • Version dropdown - Shows data layer version and timestamp
  • Entity list - All business entities and relationships between them AstroBee has identified

Entity browser

Shows information about the selected entity in two tabs:
  • Properties & Relationships
  • Data
Properties and relationships view showing entity structure
Properties section shows how database columns map to business concepts:
  • Company Domain - String type identifying company website
  • Industry Sample - The industry category (e.g. healthcare, e-commerce)
  • Company Domain Normalized - Cleaned version of domain for matching created by AstroBee
Relationships section shows how entities connect:
  • Contact via company_domain_normalized (Many to One relationship)
  • Links show the join conditions between entities

Understanding your data layer

Entities

Each entity represents a core business concept that AstroBee has identified from your data:
EntityDescription
CompanyOrganization associated to contacts, product users, and billing based on website/domain
ContactA person engaged via marketing/sales (lead, MQL, SQL, customer) from CRM
Product UserAn application user with account activity and engagement metrics
Billing AccountThe paying account profile used for invoicing and payments
SubscriptionA recurring billing plan agreement status with MRR, currency, and start dates
PlanThe commercial tier (e.g., Starter, Growth, Enterprise) associated with users and subscriptions

Properties

Properties describe each entity and come in different types:
Property TypeDescription
IdentifiersUnique keys like shipment_id or carrier_name
Descriptive fieldsAttributes like dates, names, descriptions
Calculated metricsDerived values like averages or totals
Data typesString, Integer, Timestamp, etc.

Relationships

Relationships show how entities connect to each other:
Relationship TypeDescription
One to ManyOne company has many billing accounts
Many to ManyMultiple relationships between entities

Editing your data layer

To edit your data layer, simply type your requests in natural language. There are two types of editing: quick editing and deep editing (accessible via the dropdown in the chat input).
Data Layer editing interface with chat

Quick edits

Quick edits are for fast, focused changes with minimal context required. Use these for simple changes to individual properties or relationships. These quick edits provide immediate feedback and take a few seconds at most without requiring deep thinking on the agent’s part - like renaming or deleting entities or other simple operations.

Making requests

Describe what you want to change in plain English:
  • “Rename Total Shipments to Shipments”
  • “Add customer satisfaction metrics”
  • “Create a new entity for product categories”
AstroBee processes your request and explains what it’s doing:
AI processing business model request

Making business model requests

AI reasoning about request

Detailed reasoning behind changes

The AI shows its thinking process. For example:
  • Renaming property - Explains the specific change being made
  • Business rationale - Why the change improves clarity
  • Impact analysis - How the change affects your model

Reviewing changes

After processing, you’ll see an update summary:
Data Layer update summary with changes

Data Layer update summary

Detailed data layer changes view

Detailed data layer changes view

The update shows:
  • What changed - Specific modifications made to your data layer
  • Generated entities - New business concepts AstroBee created
  • Follow-up suggestions - Recommendations for further improvements
  • Save or discard - Options to apply or reject the changes

Saving your changes

After reviewing the updates:
Data Layer saved successfully notification
  • Click “Save Model” to apply changes permanently
  • “Discard Changes” to cancel and return to the previous version
  • Success notification confirms your data layer has been updated

Deep edits

Deep edits use reasoning to research and synthesize multi-step edits. Switch to this mode via the dropdown for complex joins, creating new business entities, restructuring relationships, or when you need comprehensive analysis across multiple data sources. For example, you might type this exact query:
Create a Customer entity: A single, analytics-ready view per company across 
marketing (contacts), product (users), and billing (subscriptions/MRR). 
This will give me account-level health and revenue in one place

Deep edits in action

  1. Complex entity creation - AstroBee successfully created a new “Customer” entity that consolidates data across multiple business domains
  2. Multi-table aggregation - Generated key measures like contact counts, user engagement metrics, billing account counts, and MRR calculations
  3. Intelligent relationships - Established proper connections between the new Customer entity and existing entities (Company, Contact, Product User, Billing Account)
  4. SQL view generation - Created sophisticated SQL that aggregates data from contacts, product usage, and billing systems
  5. Business logic integration - Applied normalization rules and assumptions for data consistency across domains

Rename data layer

You can rename your entire data layer to better reflect your business or project. Click the ellipsis () next to your data layer name, select ” Rename” and enter a new name like “Sass data layer ” The change applies immediately and creates a new version, updating both the display name and system identifier to keep your data layers organized as projects evolve.

Delete data layer

To permanently delete a data layer, click the ellipsis () next to the data layer name and select ” Delete.” You’ll see a confirmation dialog warning that this action cannot be undone. Once deleted, the data layer is completely removed and you’ll be redirected to the model selection screen to choose or create a different data layer .

Version management

Your data layer is versioned automatically:
  • Timestamps show when each version was created
  • Version numbers increment with each change (v5, v6, etc.)
  • Rollback capability lets you return to previous versions if needed
Data Layer version management interface

Best practices

Be specific in your requests - Instead of “improve the model,” say “add customer retention metrics” Review changes carefully - Check that AI modifications match your business understanding Use business terminology - Describe changes using language your team understands Test after changes - Ask questions in the main interface to verify improvements work as expected

Next steps