Skip to main content

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.

This guide walks through the catalogue-first flow: create a catalogue, connect a data source, run operations, review the changes, and sync downstream. If you only need a one-off dataset, skip ahead to Research Agents.

​
Step 1 · Create a catalogue (1 minute)

  • From the Dashboard, choose Catalog Operations → Onboard & enrich SKUs (or click Start managing your catalog).
  • Name the catalogue (e.g. Products, Suppliers) and pick a starting object type. Each catalogue is opened with five tabs: Attributes, Records, Data Source, Operation, Config.

​
Step 2 · Define your attributes (1–2 minutes)

  • On the Attributes tab, add the typed fields your records need: text, number, image, date, enum, reference.
  • Mark fields as required, set enum values, and configure references between catalogues (e.g. Product.supplier → Supplier).
  • You can edit the schema later — Claro tracks attribute history alongside record history.

​
Step 3 · Connect a data source (2 minutes)

  • On the Data Source tab, pick an upstream input:
    • File upload (CSV, XLSX)
    • Scheduled scrape or HTTPS pull
    • Supplier Portal submission
    • Database connector (BigQuery, Postgres, Supabase)
  • Map the source columns to your attributes. The mapping is saved per source and reused on subsequent syncs.

​
Step 4 · Run your first operation (2–3 minutes)

Open the Operation tab. The operation grid shows everything you can run against this catalogue. A typical first run:
  1. Validate Data — surfaces schema issues, missing required fields, and rule violations.
  2. Normalize Data — standardizes units, currencies, dates, and casing across the records.
  3. Bulk Enrichment — fills missing attributes from web search, model generation, supplier docs, or another catalogue.
Each operation produces a run with a reversible diff and a confidence score per change.

​
Step 5 · Review and apply (1–2 minutes)

  • Above-threshold writes are auto-applied.
  • Below-threshold writes queue in Notifications for human review.
  • Review items can be approved, rejected, or edited inline; bulk actions are available per cluster.

​
Step 6 · Sync downstream (1 minute)

  • Under Distribute, configure Sync & Export with the connectors you need: Shopify, Amazon, BigQuery, Google Sheets, Webhooks, S3.
  • Or expose the Unified Catalog as a query-ready view for downstream consumers.

​
Tips

  • Start small — load 50–100 records, run the operations end-to-end, and tune the thresholds before scaling up.
  • Chain operations into a pipeline (Onboard → Validate → Normalize → Bulk Enrich → Push & Sync) so each new batch of records flows through the same gates.
  • Every value carries provenance — the operation, run, source, model, and reviewer that produced it. Roll-back is per-field.
Ready to go deeper? See Catalogue, Operations, and the module guides for Onboard, Analyse, Monitor, and Distribute.