> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getclaro.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Catalogue

> Your products, suppliers, SKUs, and the relationships between them — with full attribute schemas, history, and provenance.

A Catalogue is a typed object — Products, Suppliers, Line Items, Variants, etc. — that holds your records and everything that acts on them. Catalogues are persistent: they accumulate history, track provenance per field, and survive across operations and runs.

Multiple catalogues can exist in a workspace and can reference each other (e.g. `Product.supplier` → `Supplier`).

***

## The five tabs

Each catalogue opens with five tabs along the top.

### Attributes

The schema for the catalogue. Each attribute is a typed field:

* **text** — free text or constrained patterns
* **number** — integers, decimals, with optional units
* **image** — single image or gallery, with metadata (dimensions, format)
* **date** — date or datetime, with timezone handling
* **enum** — a fixed value set, with optional ordering
* **reference** — a pointer to a record in another catalogue

Per-attribute settings include: required, default value, validation rules (regex, range, length), and (for references) the related catalogue and reverse relationship name.

Edit the schema at any time — Claro tracks attribute history alongside record history. Adding or changing an attribute does not destroy existing data; type migrations produce a preview before they are applied.

### Records

The actual rows. Sortable, filterable, editable in a spreadsheet-like grid.

* Inline edits create change events with provenance.
* Per-field history is browsable and roll-back is per-field.
* Filters, saved views, and bulk edits are first-class.
* Each record has a stable identity that survives across data source updates.

### Data Source

The upstream feeds that populate the catalogue.

* **File upload** — CSV, XLSX (one-off or repeated)
* **Scheduled scrape** — URL templates with mapping
* **HTTPS pull** — periodic pulls of JSON or CSV endpoints
* **Supplier Portal** — submissions land here pre-mapped
* **Database connectors** — BigQuery, Postgres, Supabase
* **Email-as-source** — emails to a workspace address are parsed

Each source has its own mapping to the catalogue's attributes, its own schedule, and its own conflict policy (overwrite, append, write-if-empty).

### Operation

The transformations available against this catalogue. The Operations tab is the launch surface for **Export**, **Price Insights & Monitoring**, **Bulk Enrichment**, **Analyse**, **Data Source Mapping**, **Push & Sync**, **Enrich Images**, **SEO Report**, **Find Duplicates**, **Find Similarities**, **Validate Data**, **Normalize Data**, and **Generate Taxonomy**.

Operations can be run on demand, scheduled, or chained. Every run produces a reversible diff and confidence per change. See [Operations](/operations) for the full reference.

### Config

Governance for the catalogue.

* **Access** — who can read, edit, and approve.
* **Retention** — history depth, archival rules.
* **Write-back targets** — which downstream connectors are allowed.
* **Default reviewers** — who receives review queue items.
* **Auto-apply thresholds** — per-operation and per-attribute confidence cutoffs.

***

## Relationships between catalogues

Reference attributes connect catalogues. A `Product.supplier` reference to a `Supplier` catalogue carries:

* Forward navigation from product to supplier on the Records grid.
* Reverse navigation from supplier to all related products.
* Cascade rules on supplier merge or delete.
* Cross-catalogue filtering (e.g. *all products from suppliers in Germany*).

Operations can read across references — Bulk Enrichment can use values from a referenced supplier record as context, for example.

***

## History and provenance

Every value in the Catalogue carries the source of its last write — which operation, which run, which underlying data source, which model, which reviewer. This is visible inline on every cell and is the basis for:

* **Roll-back** — per-field, to any prior value.
* **Audit** — full activity trail per record.
* **Trust** — every model-generated value is traceable to its inputs.

***

## Promoting a dataset to a Catalogue

Datasets produced by Research Agents are not catalogues by default. To promote one:

1. Open the dataset in **Generated Datasets**.
2. Choose **Promote to Catalogue**.
3. Pick the target catalogue type or create a new one.
4. Map columns to attributes; new attributes can be created on the fly.

Once promoted, the records become persistent and subject to all the same operations, provenance tracking, and review gates as native catalogues.