> One semantically enriched index across DAM, PIM, ECM and custom systems: the foundation for portals and RAG. Your systems stay the system of record.

Source: https://www.smint.io/platform/uscl  
Title: The Unified Semantic Context Layer

---

# Your entire content stack.  
Consolidated into one solid platform foundation.

The Unified Semantic Context Layer brings data from your source systems into one central, semantically enriched index and serves as the foundation of the platform. Your systems remain the system of record.

[Request a technical deep dive](https://www.smint.io/technical-deep-dive)[How it works](#how-it-works)

-   [Platform](https://www.smint.io/platform)
-   [Portals](https://www.smint.io/platform/portals)
-   [Sminted AI](https://www.smint.io/platform/sminted-ai)

## How it works

The Unified Semantic Context Layer (USCL) synchronizes with your source systems in near real-time: push-based where it is supported, Smint.io pulls where it is not. On ingestion, data is consolidated, enriched and indexed, ready for every other part of the platform.

Please note that ingestion is not the only path. A live connection straight to a source system, bypassing the layer, is fully supported where a use case calls for it.

### Your existing systems

-   DAM
-   PIM
-   ECM / SharePoint
-   CMS
-   MRM
-   Custom systems

### Smint.io

**Unified Semantic Context Layer (USCL)**

One central, semantically enriched index. Your systems stay the system of record.

-   ConsolidateMap data onto one shared semantic model
-   EnrichPreviews, renditions, metadata extraction, embeddings, search relevance
-   IndexOne search and vector index across all sources

### Portals

The experience for humans

### Sminted AI

The experience for AI

### API consumers

Your own applications

## Three jobs, one layer

Three things have to happen to your content before people, AI and applications can use it. The context layer does all three for you, three projects you don't have to build yourself.

### Consolidate

Data from many source systems is mapped onto one shared semantic model; the layer understands what an asset, product, rendition or collection is. The result is a central search and vector index across heterogeneous sources.

Many access points → one access point

### Enrich

Preview and media formats, streaming renditions, metadata extraction (XMP, EXIF), embeddings for semantic and visual search, and a relevance layer that tunes itself to your data. Data comes out better than it went in.

Hard to use → easy to use

### Enable

The layer shields your source systems from live traffic, effectively lifts hard API quotas, and keeps portals running even when a source system is down.

Technically impossible → possible

## The physics of IT

Some constraints in IT behave like physics, and they apply to AI too. The USCL enables AI by lifting hard constraints that no level of intelligence can work around.

**Intelligence**

Subject to the physics of IT.

-   **Indexing & relevance**Searching a library without an index takes ages at any intelligence level. Irrelevant search results break any logic.
-   **API quotas**A hard API quota is just as hard for AI, and agent traffic reaches it far sooner.
-   **Load**A source system that collapses under live traffic collapses under agent traffic even faster.
-   **Context limits**A context window that is full is full. Wasting tokens costs significant amounts of money.
-   **Aggregation**Data scattered across many systems must be aggregated by _someone_ before anyone, human or model, can reason over it.

## Search results that are right the first time

The relevance layer reads the data it indexes and tunes search to it, automatically, for your content, your languages and your naming conventions. The right asset comes first, for a person typing a query and for an AI assembling context.

In RAG terms: the governed retrieval layer. One continuously synchronized corpus with embeddings and semantic models, permission-aware at query time.

## Composable assets

A composable asset is a record that lives in Smint.io and references content in your systems. Its structure is free, so it can be a press release, a product page, a campaign or anything else your audiences need as one thing.

A press release, for instance, holds headline, text and other metadata on the platform, and references illustrations in the DAM or download files in the ECM. It is found, permissioned and distributed like any other asset, and it stays up-to-date as the referenced parts change.

## One permission model for people, AI and applications

Who may see what, in which market, under which licence, until which expiry: decided once, and the same for a portal user, an AI client via Sminted AI and an application via the API.

-   Metadata-based permissions via Dynamic Content Routing, across all connected systems.
-   Pass-through of your source systems' own permissions where they fit.
-   Audience-specific scopes for consumers your source systems were never set up for.
-   SSO via SAML/OIDC with SCIM provisioning.
-   Request and approval workflows.
-   Full audit trail for human and AI access alike.

## How it fits your stack

### What is stored

The search index, metadata and previews. Originals stay in your systems and are fetched on demand, or fully synchronized where a use case needs it. Your management systems remain the system of record; there is no data migration project.

### With your middleware

Your data layer already consolidates. Smint.io docks onto it, refines the data and adds the experiences your middleware does not provide. _In production at customers that built their own data layer on purpose._

### As your middleware

No usable data layer in place? The USCL becomes it: the one basis for portals, Sminted AI and your own applications. _Arper runs its DAM and its PIM on it; other customers supply partners with data through the Smint.io API alone._

### Fed by 20+ connectors

DAM, PIM, ECM, MRM and templating systems out of the box; custom and in-house systems via the connector SDK.

-   DAM
-   PIM
-   ECM
-   MRM
-   Templating
-   Custom systems

[All supported systems](https://www.smint.io/platform/connectors)

## Talk architecture with the people who built the layer

Bring your stack diagram. We'll show you where the layer sits, what it stores, and how it behaves when a source system doesn't.

[Request a technical deep dive](https://www.smint.io/technical-deep-dive)
