How a catalog becomes discovery.
What the engine does, what integration requires of you, and what is live today versus in build.
Connect. Orchestrate. Launch.
Your catalog comes in as it is.
Metadata is normalized into the shared schema. Audio, text, image, and structure are ingested by the multimodal stack.
The catalog becomes a graph.
Each node is backed by its own vector store and structured by an ontology built for your domain. Model-agnostic orchestration routes tasks across foundation models.
Discovery ships inside your product.
Semantic search, conversational discovery, and taste intelligence, measured by the evaluation harness.
What the engine is made of.
| Layer | What it is | Why it matters |
|---|---|---|
| Knowledge graph | Domain entities as graph nodes, each backed by a dedicated vector store. | Retrieval by meaning, grounded in structure. |
| Domain ontologies | A schema of the vertical, built per deployment. | Results read as native to the catalog. |
| Metadata schema | One normalization layer shared by every deployment. | New catalogs onboard against a known shape. |
| Multimodal stack | Ingestion and matching across audio, text, image, and structure. | The catalog is searchable by what it contains, not how it is labeled. |
| Taste profiling | Learns preference by contrast, sharp on day one. | No cold start. No behavior history required. |
| Evaluation harness | Relevance measured against outcomes on every deployment. | Quality is tested, not asserted. |
| Orchestration | Model-agnostic routing across foundation models. | Better models lower our cost and raise our quality. |
Four ways in.
Each path states what it requires of you.
Discovery API
Direct integration into your own surfaces. Requires your engineering against the API and a catalog feed.
Embeddable component
A drop-in discovery component inside your pages. Requires a script include and a container element.
Iframe overlay
The fastest path. Requires one embed snippet. No changes to your stack.
White-label standalone
A discovery surface under your brand. Requires a catalog feed and a domain.
Live today. In build.
Today each deployment is configured per customer. Here is exactly what that means.
Live today
- Taste intelligence, semantic search, conversational discovery in production.
- Per-customer configured deployments.
- Multimodal ingestion across audio, text, image, and structure.
- Evaluation harness on every deployment.
- One paid deployment delivered on time and on scope: Minnesota Star Tribune.
- Sub-2-second response across 10,000+ live searches. No outage.
In build
- Unified, versioned Discovery API.
- Tenant isolation.
- Usage metering.
The gate that defines repeatable: one engineer stands up a new media partner in under ten working days.
Stated plainly.
Isolation
Deployments are isolated per customer today. Multi-tenant isolation ships with the Discovery API.
Metering
Usage metering is in build and arrives with the API.
Insurance
Full commercial liability, E&O, and cyber insurance through The Hartford.
Certifications
We hold no security certifications and claim none.
Contact
See it on your catalog.
The fastest way to evaluate the engine is a working demo on your own catalog. Tell us what your audience cannot find.