Datasources

Each datasource can have its own context.

This lets Querio understand one datasource differently from another, even when the same business term appears in both places.

Catalog

The catalog is a JSON-backed map of your datasource.

Querio turns it into UI tabs so it is easier to navigate, but conceptually it is one structured catalog file.

The catalog includes:

• Schemas
• Tables
• Columns
• Joins
• Descriptions
• Agent access controls

Schemas

Schemas group tables inside a datasource.

Use schema descriptions to explain the broad purpose of a schema and when Querio should use it.

Tables

Tables are the core objects Querio queries.

Use table descriptions to explain what the table represents, what grain it has, and when it should or should not be used.

Columns

Columns describe the fields inside a table.

Use column descriptions to explain business meaning, expected values, and any caveats.

Joins

Joins explain how tables relate to each other.

Good join context helps Querio avoid incorrect joins and makes multi-table analysis more reliable.

Knowledge Base

The Knowledge Base contains context Querio reads to understand your business and data, as well as reusable building blocks.

When you create or edit a file in the Knowledge Base, you fill in three fields:

• Title — a short name for the file
• Description — what this file covers and when it is useful
• Context type — either Always or Reference

Always means Querio loads this file into every conversation automatically.

Reference means Querio only uses it when it decides the content is relevant.

Use Always for core definitions and rules your team applies constantly. Use Reference for things that apply to specific topics or edge cases.

Examples of what to put in the Knowledge Base:

• Metric definitions
• Business rules
• Data caveats
• SQL examples
• Python examples
• Team-specific analysis notes

Frontmatter

Each Knowledge Base file starts with frontmatter that controls how Querio uses it. The format depends on the file type:

• .md files use a --- block
• .sql files use -- field: value lines
• .py files use # field: value lines

Three fields are supported:

• name — a short title for the file
• description — one-line summary of what the file covers
• context — either always or reference

context: always means the full file body is included in Querio's prompt on every turn. Use this for rules, metric definitions, naming conventions, and hard constraints that must apply universally.

context: reference means only the file's name and description appear in the prompt index. Querio fetches the full content when a question matches. Use this for query templates, SQL examples, and situational context that only applies sometimes.

New files in the Knowledge Base default to context: reference to keep the prompt lean. Promote to context: always only when the content needs to apply on every single turn.

How Querio saves context for you

When you share a fact, correction, or hard-won query during an analysis, Querio will offer to save it to the Knowledge Base at the end of its response. It names the destination file and waits for your approval before writing anything.

You can decline. If you do, it won't offer again in that session.