Knowledge

Good analysis isn't just about writing SQL—it's about understanding how your business relates to the data you capture.

This section is where you teach Qluent about your business:

• Which data should Qluent have access to? Enable/disable tables and columns
• What does each table, column and value represent? Add descriptions and enable semantic search
• What background does Qluent need? Write instructions to provide context

The more accurate your configuration, the better Qluent's answers will be.

Illustrates adding descriptions to tables and columns in the data model.

Instructions

Think of instructions as the guide you'd give a new colleague when onboarding them to your data. What context do they need before they start writing queries?

Instructions provide the background—what the business does, where the data comes from, and how your team talks about things. Then use metrics and validated queries to get the specific calculations and query patterns right.

Use cases

• What your company does and how it makes money
• Where the data comes from and how often it's updated
• Terminology your team uses that might differ from column names
• General context that helps interpret the data correctly

Example

# About our business
We're an e-commerce company selling electronics. Our main markets are US and Europe.
The fiscal year starts in April. When someone asks about "this year" they usually mean fiscal year.

# Our data
- Orders come from our Shopify store and are synced daily
- We also have marketing data from Google Analytics and Meta Ads
- Customer support tickets are in a separate Zendesk table

# Terminology
- "Customers" means people who have completed at least one order
- "Users" includes everyone who created an account, even without orders
- "Products" and "SKUs" are used interchangeably here

# Things to watch out for
- Test orders exist with customer_id = 0
- Prices are stored in cents
- NULL in the country column means "Unknown"

For specific calculations like revenue formulas, use metrics. For specific query patterns, use validated queries.

Data Model

Tables and columns available to Qluent are listed here. They are automatically detected based on the schema of your connected data source. You can:

• Enable/disable each table and column
• Set a description for each table and column
• Enable semantic search on columns

Why descriptions matter

When a user asks "What were our top-selling products last month?", Qluent needs to know which table contains sales data, which column has product names, and what "top-selling" means.

Without good descriptions, Qluent has to guess—and guesses lead to wrong answers.

Example descriptions

Tables:

• ❌ Poor: orders → "Orders table"
• ✅ Good: orders → "Customer orders. Each row is one order. Only includes completed orders (not cancelled)."

Columns:

• ❌ Poor: status → "Status"
• ✅ Good: status → "Order status. Values: 'pending', 'processing', 'shipped', 'delivered', 'cancelled'. Use 'delivered' for completed orders."

Tips

• Disable internal IDs, audit columns, and deprecated fields
• List possible values for status/type columns
• Explain relationships: "Links to customers table via customer_id"
• Note data quirks: "Amounts are in cents, divide by 100 for dollars"

Tables and columns prefixed with underscores (e.g. _my_private_table) are ignored by default.

Getting started

Recommended order for configuring a new project:

1. Build the data model - Connect your data source, disable irrelevant tables/columns, write descriptions
2. Add instructions - Write the onboarding guide for your data
3. Enable semantic search - Turn on semantic search for categorical columns
4. Create metrics and examples - Add metrics for key calculations and validated queries for query patterns