Knowledge Base - Superatom AI

The Knowledge Base is where Superatom connects to your data and builds understanding of your organization. It includes data sources, file management, integrations, and knowledge graphs.

Data Sources

Connect to your enterprise databases and data warehouses:

Connected Data Sources — Connected data sources overview

Data Sources Tab — Data sources tab — connect to databases, warehouses, files, APIs, CRM, or cloud

Supported Databases

PostgreSQL

Open-source relational database

MySQL

Popular relational database

SQL Server

Microsoft enterprise database

BigQuery

Google Cloud data warehouse

Snowflake

Cloud data platform

ClickHouse

Column-oriented database

Databricks

Unified analytics platform

Redshift

AWS data warehouse

Connection Types

Connect Data Warehouses — Data Warehouses

Connection Times

Source Type	Typical Setup Time
PostgreSQL, MySQL, SQL Server	< 1 hour
BigQuery, Snowflake, Redshift, Databricks, ClickHouse	< 1 hour
SAP, Oracle ERP, Dynamics 365	1-3 days
Excel, CSV, Parquet files	< 30 minutes
REST APIs, GraphQL	1-2 hours

Connecting a Database

Click Add Connection

Navigate to Knowledge Base → Data Sources → Add Connection

Select Database Type

Choose from the supported database types

Add Data Source Form — Add a new data source connection

Enter Connection Details

Host/Server address
Port number
Database name
Username and password (read-only)

Test Connection

Verify connectivity before saving

Automatic Semantic Modeling

Superatom runs a five-stage automated analysis to build a semantic model from your data

Automated Semantic Modeling

When a data source is connected, Superatom automatically builds a semantic model through five stages:

Schema Analysis

Reads database structure: tables, columns, data types, relationships, foreign keys, indexes

Sample Data Extraction

Examines representative samples to understand actual data characteristics, edge cases, and formats

Statistical Profiling

Calculates distinct values, distributions, frequencies, ratios, and patterns across columns

Domain Classification

Recognizes industry context (supply chain, retail, finance, healthcare) and applies domain-specific mappings

Question Generation

Proposes common, diagnostic, and exploratory questions that can be asked of this data

Traditional semantic modeling requires specialized consultants and takes 3-6 months. This automated process completes in approximately 2 days and improves continuously as the system is used.

ERP Systems

Connect directly to enterprise resource planning systems:

Supported ERPs

System	Description
SAP	Integrated business processes and real-time data
Oracle ERP Cloud	Complete suite of cloud applications
Microsoft Dynamics 365	CRM and ERP capabilities

Files & Documents

Upload and manage reference documents:

Document Types

Data Dictionary

Reference of tables, columns, and business definitions

Business Glossary

Key metrics and KPI definitions

Analytics Guide

Best practices for customer segmentation, cohort analysis

Schema Documentation

Entity relationships and data flow

Documents are:

Searchable — AI uses them for context
Categorized — By type and topic
Versioned — Track changes over time

Integrations

Connect to third-party tools and services:

Available Integrations

Gmail
Outlook

Knowledge Graphs

Visualize and manage organizational knowledge:

What Knowledge Graphs Show

The graph visualizes relationships between:

Entities — Orders, Customers, Products, etc.
Processes — Order lifecycle, payment flow, shipping
Categories — Overview, Structure, Data, Relationships, Insights

Node Categories

Category	Color	Description
Overview	Blue	High-level concepts
Structure	Green	Organizational structure
Data	Yellow	Data entities
Relationship	Pink	Entity connections
Process	Red	Business processes
Insight	Purple	Derived knowledge

Knowledge Nodes

Create and manage knowledge nodes that guide AI analysis:

Knowledge Nodes Page — Knowledge nodes management page

Node Types

Type	Purpose
Global	Organization-wide rules and definitions
User	Role-specific or personal preferences
Query	Definitions for specific questions

Creating a Knowledge Node

Knowledge Node Detail - Overstock Definition

Create Knowledge Node — Create a new knowledge node

Edit Knowledge Node — Edit an existing knowledge node

Each node includes:

Title — Clear, descriptive name
Description — What this knowledge represents
Rules — Specific conditions and calculations
Key Columns — Data fields involved
Formula — How to calculate (if applicable)

Why Tribal Knowledge Matters

Accuracy

Answers reflect how your organization actually operates, not just what the raw data says

Consistency

The same term means the same thing for every user across the organization

Onboarding

New employees get organization-aware AI from day one — no ramp-up period

Knowledge Preservation

Institutional knowledge persists through employee turnover instead of walking out the door

Example: Overstock Definition

title: "Overstock"
type: "classification"

description: |
  Overstock items are items where the On Hand Cash Value
  exceeds the Cost of Goods Sold (COGS) over a rolling period.
  The rolling period varies by product type.

rules:
  - "Structural products (ProductType = 'Structural'): Rolling 90 days COGS > $0"
  - "Decking products (MajorMinorCode = '0103DEC'): Rolling 180 days COGS > $0"
  - "Specialty products: Rolling 120 days COGS > $0"

key_columns:
  - "FactInventoryDaily.OnHandInventory"
  - "FactSalesInvoice.AverageCostExtended"
  - "DimItem.ProductType"
  - "DimItem.MajorMinorCode"

formula: "OnHandInventory - (Rolling Period COGS)"

Data Model

View the semantic model Superatom has built:

Edit Semantic Model — Edit the semantic model

What’s Included

Tables — All discovered tables and their purposes
Columns — Field-level details with data types
Relationships — Foreign keys and joins
Statistics — Row counts, distributions, patterns

Using the Data Model

The data model powers:

Natural language understanding — Mapping questions to tables
Query generation — Building correct SQL
Visualization selection — Choosing appropriate charts
Tribal knowledge application — Applying org-specific rules

Best Practices

Start with Core Data

Connect your most critical data sources first. Add more as needed.

Document Business Terms

Create knowledge nodes for every important business term.

Keep Documents Updated

Upload current data dictionaries and business glossaries.

Review the Graph

Periodically review the knowledge graph for accuracy.

Next Steps

Tribal Knowledge

Deep dive into the knowledge system

Semantic Modeling

How Superatom understands your data

​Data Sources

​Supported Databases

PostgreSQL

MySQL

SQL Server

BigQuery

Snowflake

ClickHouse

Databricks

Redshift

​Connection Types

​Connection Times

​Connecting a Database

​Automated Semantic Modeling

​ERP Systems

​Supported ERPs

​Files & Documents

​Document Types

Data Dictionary

Business Glossary

Analytics Guide

Schema Documentation

​Integrations

​Available Integrations

​Knowledge Graphs

​What Knowledge Graphs Show

​Node Categories

​Knowledge Nodes

​Node Types

​Creating a Knowledge Node

​Why Tribal Knowledge Matters

Accuracy

Consistency

Onboarding

Knowledge Preservation

​Example: Overstock Definition

​Data Model

​What’s Included

​Using the Data Model

​Best Practices

​Next Steps

Tribal Knowledge

Semantic Modeling

Data Sources

Supported Databases

Connection Types

Connection Times

Connecting a Database

Automated Semantic Modeling

ERP Systems

Supported ERPs

Files & Documents

Document Types

Integrations

Available Integrations

Knowledge Graphs

What Knowledge Graphs Show

Node Categories

Knowledge Nodes

Node Types

Creating a Knowledge Node

Why Tribal Knowledge Matters

Example: Overstock Definition

Data Model

What’s Included

Using the Data Model

Best Practices

Next Steps