Skip to main content

​
Overview

A Corvic Table is a distributed processing construct that defines what entity, object, or operational unit you want to work with across your data. Rather than being tied to a single data source or format, a Corvic Table composes multiple contributing inputs—structured, unstructured, or multimodal—into a unified, addressable representation. Corvic Tables serve as the backbone for distributed data processing and transformation. They let you declaratively specify how complex data should be joined, enriched, normalized, and evolved at scale—without hard-coding brittle pipelines. Each table becomes a logical unit for computation, enabling scalable transformations, feature generation, operational analytics, and downstream AI workflows. By abstracting what the entity is from how the data is physically stored or processed, Corvic Tables allow teams to build reusable, composable intelligence that runs efficiently across distributed data infrastructure.
A Corvic Table specifies what needs to be converted to embeddings, but it doesn’t explain how it should be embedded. That’s the role of a Space.

​
Key Concepts

​
Entities

Entities are the objects you want to embed. Examples include:
  • Account IDs
  • Customer profiles
  • Product information
  • Document identifiers

​
Input Sources

Input sources are the data sources that contribute to defining your entities:
  • Entity (dimension) files: Direct entity information
  • Relation (fact) files: Relationships between entities

​
Space Keys

Space keys are the unique identifiers for entities in your embedding space:
  • Typically an ID field from your entity files
  • Must be unique across all entities in the space
  • Used to reference entities in embeddings

​
Creating a Corvic Table

​
Step 1: Select Input Sources

Choose the data sources that will contribute to your Corvic Table:
  1. Navigate to Corvic Tables in your data room
  2. Click “Create Corvic Table”
  3. Select one or more input sources
  4. Review the schema of selected sources

​
Step 2: Define Entities

Specify the entities you want to embed:
  1. Select the entity type (e.g., Account, Customer, Product)
  2. Choose the space key field (e.g., account_id, customer_id)
  3. Map fields from input sources to entity attributes

​
Step 3: Configure Relationships

If using relation files, configure how they connect entities:
  1. Select relation files
  2. Define relationship types
  3. Specify how relationships contribute to embeddings

​
Step 4: Save

  1. Provide a descriptive name
  2. Add an optional description
  3. Review your configuration
  4. Save the Corvic Table

​
Example: Financial Accounts

Suppose you have:
  • accounts.parquet: Account information with account_id
  • transactions.parquet: Transactions between accounts
You can create a Corvic Table for accounts:
  • Input Sources: accounts.parquet, transactions.parquet
  • Entity: Account
  • Space Key: Account.account_id
  • Relationships: Transactions link accounts together
This Corvic Table defines accounts as entities to embed, using both account information and transaction relationships.

​
Corvic Table Types

​
Single Entity

Embed a single entity type:
  • One entity file
  • One space key
  • Simple structure

​
Multi-Entity

Embed multiple entity types:
  • Multiple entity files
  • Multiple space keys
  • More complex relationships

​
Graph-Based

Embed entities with rich relationships:
  • Entity and relation files
  • Complex relationship structures
  • Graph structural embeddings

​
Best Practices

​
Entity Selection

  • Choose entities that are meaningful for your use case
  • Ensure entities have sufficient data
  • Consider entity relationships

​
Input Source Selection

  • Include all relevant data sources
  • Ensure data quality in input sources
  • Consider data freshness

​
Space Key Selection

  • Use unique, stable identifiers
  • Avoid keys that change over time
  • Ensure keys exist in all relevant input sources

​
Validation

Before saving, the platform validates:
  • Space key uniqueness
  • Input source availability
  • Schema compatibility
  • Relationship definitions

​
Managing Corvic Tables

​
Viewing Corvic Tables

  • List all Corvic Tables in a room
  • View Corvic Table details
  • Preview entity counts
  • Check dependencies

​
Updating Corvic Tables

  • Modify input sources
  • Update entity definitions
  • Adjust relationships
  • Rename or delete
Updating a Corvic Table may affect dependent spaces. Review dependencies before making changes.

Spaces

Create spaces from Corvic Tables to generate embeddings.

Data Sources

Learn about uploading data sources.

Sample Datasets

See examples of Corvic Tables in sample datasets.