> ## 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.

# Advanced Features

> Dedicated-plan capabilities for high-volume, enterprise-grade catalog operations.

The features on this page are available on Dedicated plans. They extend the core platform for organizations running catalogs at scale, with requirements around custom models, real-time predictions, human feedback loops, and bespoke integrations.

***

## Taxonomy generation & dynamic ontology

Generate sophisticated multi-level taxonomies from your data and keep them evolving as your catalog grows.

* **Multi-level hierarchy** — 3–8 levels deep based on data complexity, auto-detected.
* **Semantic relationship mapping** — relationships between nodes are explicit and queryable.
* **Dynamic ontology evolution** — as new data patterns emerge, Claro proposes taxonomy extensions for review.
* **Cross-domain merging** — consolidate taxonomies from multiple catalogs or data sources into a unified structure.
* **Industry-standard templates** — GS1, Google Product, UNSPSC as starting points.
* **Validation** — automated consistency checking flags contradictions, orphaned nodes, and coverage gaps.

For the standard (non-Dedicated) version, see [Taxonomy](/taxonomy) and the [Generate Taxonomy operation](/operations#generate-taxonomy-beta).

***

## Entity deduplication & resolution

Enterprise-grade entity matching that goes beyond string comparison.

* **Fuzzy string matching** — configurable similarity thresholds per field.
* **Phonetic matching** — handles name variations and misspellings across locales.
* **ML-powered semantic similarity** — embedding-based matching for product names, descriptions, and addresses.
* **Cross-field correlation** — match entities across different attribute combinations, not just a single key.
* **Probabilistic record linkage** — scored match proposals with explicit confidence.
* **Negative-example learning** — rejected pairs are remembered and suppress similar proposals.
* **Custom matching rules** — domain-specific business logic (e.g. match on GTIN first, fall back to brand + MPN).

For the standard version, see [Similarity & Duplicate](/similarity_duplicate) and [Find Duplicates](/operations#find-duplicates-beta).

***

## Near real-time predictions (\< 200 ms)

Ultra-low-latency prediction endpoints for production environments where speed is critical.

* **Dedicated GPU clusters** with model caching — no cold start.
* **Edge deployment** — geographic optimization for latency-sensitive use cases.
* **Auto-scaling** — matches request volume without provisioning.
* **SLA guarantees** — 99.9% uptime with real-time monitoring and alerting.
* **Load balancing** across multiple regions.

**Use cases:** Live product recommendations, real-time content moderation, instant classification at point of entry, dynamic pricing feedback.

***

## Human-in-the-loop feedback system

Systematic expert review that makes models continuously better over time.

* **Review interface** — domain experts approve, correct, or flag AI outputs with an annotation tool.
* **Confidence calibration** — expert agreement is tracked and used to refine scoring algorithms.
* **Automated retraining** — daily (or configurable: hourly / weekly) model updates incorporating all accumulated feedback.
* **Performance tracking** — dashboards showing improvement over time per operation and per attribute.
* **Multi-expert consensus** — track agreement across reviewers for contested cases.
* **A/B testing framework** — compare model versions before promoting to production.
* **Expert validation** — optional gating of model updates behind expert sign-off before deployment.

For standard review workflows, see [Notifications](/notifications).

***

## Advanced API integration hub

Connect Claro to proprietary datasets and specialized external services.

* **Proprietary dataset access** — direct connections to internal databases, data lakes, and enterprise systems.
* **Domain-specific APIs** — industry databases, patent systems, regulatory filings, ERP and PIM solutions.
* **Auth management** — OAuth, API keys, certificate-based, and custom schemes.
* **Data transformation pipelines** — ETL for complex format conversions before values land in the catalog.
* **Intelligent caching** — responses are cached to reduce cost and improve latency on repeated lookups.
* **Rate-limit management** — quota handling across multiple providers, with backoff and retry.

For standard inbound and outbound connectors, see [Data Import & Ingestion](/data_import_ingestion) and [Distribute](/modules/distribute).

***

## Custom analytics

A BI suite surfacing catalog health, operation performance, and team activity across the workspace.

* Interactive dashboards — configurable per team and per catalog.
* Real-time monitoring of operation run metrics.
* Data quality trends over time.
* Credit consumption and efficiency breakdowns.
* Export to BigQuery or BI tools via Webhooks.

***

## Self-hosting

Run Claro in your own infrastructure:

* **Kubernetes Helm chart** — deploy to any cluster.
* **Managed VPC** — Claro sets up a dedicated environment in your cloud account.
* **Air-gapped** — fully isolated deployment with no external calls.
* **On-premises** — bare-metal or private cloud.

Contact your success team for deployment architecture and sizing guidance.

***

## Getting access

All features on this page require a Dedicated plan. Talk to us at **[hello@getclaro.ai](mailto:hello@getclaro.ai)** or schedule a call from the dashboard. We'll scope the right configuration for your catalog volume and use case.