> ## 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.

# Operations

> The verbs of the platform. Every transformation is an auditable run with confidence scores and a reversible diff.

Operations are the building blocks of the Catalog Operations workflow. Every operation runs against a catalogue (or a filtered subset), produces a reversible change set, and writes scored, traceable updates back to records.

This page is the reference for every operation currently available. Find them under the **Operation** tab on any catalogue.

***

## How operations run

An operation always:

1. **Reads** records (the entire catalogue, or a saved filter).
2. **Optionally calls** models, web sources, or another catalogue.
3. **Produces a run** — a record of inputs, outputs, and per-change confidence.
4. **Writes back** changes, gated by the auto-apply / review / reject thresholds you've configured.

Operations can be:

* **Run on demand** — single execution against a filter.
* **Scheduled** — recurring (hourly, daily, weekly, on-event).
* **Chained** — output of one operation triggers the next (e.g. *Onboard → Validate → Normalize → Bulk Enrich → Push & Sync*).

Every run is reversible. The catalogue maintains per-field history; rolling back affects only the field(s) you select.

***

## Operation reference

| Operation                       | What it does                                                                                                                        |
| ------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------- |
| **Export**                      | Produce a file in CSV, XLS, Shopify, or Amazon format. Filters and column selection are honored.                                    |
| **Price Insights & Monitoring** | Configure recurring price checks across marketplaces; pricing trends and anomalies feed into Monitor → Price.                       |
| **Bulk Enrichment**             | Fill missing or incomplete attributes at scale. Sources can be web search, model generation, supplier lookup, or another catalogue. |
| **Analyse**                     | Run the data-quality report (per-field scores, distribution stats, critical issues, overall score).                                 |
| **Data Source Mapping**         | Match uploaded files to existing records using text similarity scoring; produces a mapping draft you can review before merge.       |
| **Push & Sync**                 | Push records to Shopify or Amazon (Beta). Maps catalogue attributes to platform fields per a configurable schema.                   |
| **Enrich Images**               | Generate or repair images with metadata: alt text, primary subject tags, dimensions, format normalization (Beta).                   |
| **SEO Report**                  | Score listings on SEO health and produce optimization suggestions per field (Beta).                                                 |
| **Find Duplicates**             | Identify duplicate records using AI matching; produces merge proposals you can approve in bulk (Beta).                              |
| **Find Similarities**           | Group similar records into clusters and surface relationships; useful for variant grouping and supplier match (Beta).               |
| **Validate Data**               | Run schema, format, and rule checks; produce an issue list with severity (Beta).                                                    |
| **Normalize Data**              | Standardize formats — units, currencies, dates, casing, language — across selected fields (Beta).                                   |
| **Generate Taxonomy**           | Propose a hierarchical taxonomy from your records. Output goes to the Taxonomy surface for curation and versioning (Beta).          |

***

### Export

Produce a file in CSV, XLS, Shopify-ready, or Amazon-ready format.

* **Inputs** — catalogue or filter, column selection, output format.
* **Outputs** — a downloadable file, with a record of the export in run history.
* **Formats** — generic CSV/XLS, plus marketplace-shaped templates that match required columns and value formats for direct upload.

### Price Insights & Monitoring

Configure recurring price checks across marketplaces. Detected drifts and anomalies feed into [Monitor → Price](/modules/monitor).

* **Inputs** — catalogue, target SKUs, source marketplaces, cadence, alert rules.
* **Outputs** — price history time series, drift alerts, pricing dashboards.

### Bulk Enrichment

Fill missing or incomplete attributes at scale. The most-used operation in most workflows.

* **Inputs** — catalogue or filter, target attribute(s), source (web search, model generation, supplier lookup, another catalogue, Knowledge Base), prompt and grounding, confidence thresholds.
* **Outputs** — proposed values per record with confidence and citations. Auto-applied above threshold; queued in Notifications below.
* **Tips** — start narrow (one attribute, 100 records), tune the prompt and threshold, then expand.

### Analyse

The data-quality report described in [Modules → Analyse](/modules/analyse).

* **Inputs** — catalogue or filter.
* **Outputs** — per-field scores, distribution stats, critical issues with one-click follow-on operations, overall Data Quality Score.
* **Side effects** — none. Analyse never modifies records.

### Data Source Mapping

Match the rows in an uploaded file (or any new Data Source) to existing records using text similarity scoring before merging.

* **Inputs** — Data Source, target catalogue, fields to drive matching, threshold.
* **Outputs** — a mapping draft: matched, unmatched, ambiguous. Review before merging.
* **Use** — first stage of every onboarding pipeline; keeps duplicates out of the catalogue.

### Push & Sync (Beta)

Push records to Shopify or Amazon. Maps catalogue attributes to platform fields per a configurable schema.

* **Inputs** — catalogue or filter, target connector, mapping, conflict policy.
* **Outputs** — a run with success/failure counts and per-record errors. Errors include the platform-side response.

### Enrich Images (Beta)

Generate or repair images with metadata.

* **Inputs** — catalogue or filter, image attribute, target metadata (alt text, subject tags, dimensions, format).
* **Outputs** — alt text, structured tags, normalized image variants, with confidence per generated value.

### SEO Report (Beta)

Score listings on SEO health and produce per-field optimization suggestions.

* **Inputs** — catalogue or filter, target attributes (title, description, meta, alt text), keyword sources.
* **Outputs** — per-record SEO score, per-field recommendations, opportunity list.

### Find Duplicates (Beta)

Identify duplicate records using AI matching; produce merge proposals.

* **Inputs** — catalogue, fields to weight, similarity threshold.
* **Outputs** — clusters of likely duplicates with per-pair confidence; merge proposals you can approve in bulk via [Similarity & Duplicate](/similarity_duplicate).

### Find Similarities (Beta)

Group similar records into clusters; useful for variant grouping and supplier match.

* **Inputs** — catalogue, fields to weight, threshold.
* **Outputs** — clusters with intra-cluster similarity scores and visualizations.

### Validate Data (Beta)

Run schema, format, and rule checks against records.

* **Inputs** — catalogue or filter, ruleset (built-in plus custom rules).
* **Outputs** — an issue list with severity, attribute, and a suggested fix where applicable.

### Normalize Data (Beta)

Standardize formats — units, currencies, dates, casing, language — across selected fields.

* **Inputs** — catalogue or filter, target fields, normalization rules (currency target, unit system, date format, casing).
* **Outputs** — proposed normalized values with provenance back to the original.

### Generate Taxonomy (Beta)

Propose a hierarchical taxonomy from your records. Output goes to the [Taxonomy](/taxonomy) surface for curation and versioning.

* **Inputs** — catalogue, fields used for clustering, target depth, optional seed taxonomy.
* **Outputs** — a draft taxonomy with proposed assignments per record. Curate, then publish a version.

***

## Confidence, review, and rollback

Three settings govern how aggressive each operation is:

| Setting                  | Effect                                                 |
| ------------------------ | ------------------------------------------------------ |
| **Auto-apply threshold** | Confidence above which a write happens without review. |
| **Review threshold**     | Below auto-apply but above this — queues for review.   |
| **Reject threshold**     | Below this — discarded with a logged reason.           |

Defaults are conservative. Per-operation, per-attribute, and per-source overrides are supported.

Every write is reversible. The catalogue maintains a per-field history; rolling back affects only the field(s) you select, and a roll-back is itself a logged event.

See [Confidence Scoring](/confidence_scoring) for how the score is computed.

***

## Chained pipelines

Operations can be chained — the completion of one triggers the next. A typical onboarding pipeline:

```
Data Source Mapping
        ↓
Validate Data
        ↓
Normalize Data
        ↓
Bulk Enrichment
        ↓
Push & Sync
```

Each stage gates the next, so failures and below-threshold items in any stage halt the pipeline (or branch to review) before the catalogue is touched.