Teach Querio about your data

Querio gets better when you teach it how your data and business work.

The Context page is where this knowledge lives. Everything on it is stored as files in your workspace, and the page shows them as views that are easy to inspect and edit.

What context is for

Context helps Querio understand:

• What tables and columns mean
• How tables join together
• Which metrics matter
• How your team defines business terms
• What reusable logic or skills the agent should follow

Main types of context

The sidebar on the Context page is organized into three sections:

• Datasources: Your connected databases. Pick one to work on its context.
• The selected datasource:
  • Tables: The catalog, one schema at a time. This is where you describe schemas, tables, and columns.
  • Knowledge: SQL, Python, and Markdown entries that define metrics, rules, and business logic tied to this datasource.
  • Joins: How this datasource's tables relate to each other. Click a join to open its table.
  • Catalog file: The whole catalog for this datasource as one JSON file. You can edit descriptions and visibility here; names, types, and joins are read-only.
• Agent context:
  • Memory: Durable workspace-level notes (MEMORY.md) that Querio can reference across all conversations.
  • Skills: Reusable instructions (e.g., compare-periods, data-dictionary) that teach Querio exactly how to perform specific, repeatable tasks.
  • Files: Everything else in your workspace, such as scripts and data files. Open a text file to edit it, or download any file from its menu.

Saving changes

Workspace settings currently support only Canonical file mode. Context and execution environments use the shared workspace files directly; there is no separate changeset publication step.

Edits on the Context page are collected into one set of unsaved changes. Use Review changes to see every change before and after, and undo individual ones. Click Save to apply them to the shared files.

Catalog description and visibility edits use revision-checked file writes. If the catalog file changes between the save's read and write, the save fails and your edits remain unsaved rather than overwriting the newer file.

If you leave the page with unsaved edits, Querio asks before discarding them.

The goal is to make Querio less dependent on one-off request detail and more aware of the way your team actually works.