Govern the data you already embedded.
Adding Gateco to a store that already holds data? Retroactive registration finds unmanaged vectors and brings them under policy without re-ingestion.
Available for any connector that supports listing. Pairs with classification suggestions to label the newly registered resources.
Why it matters
A vector you have not registered is a vector no policy can protect. When Gateco is introduced to an existing pipeline, the data already in the store is invisible to the policy engine until it is registered as a gated resource. Retroactive registration closes that gap: it walks the connector, finds the vectors Gateco does not yet know about, and registers them so they are evaluated like everything else.
This is as much a discovery tool as a setup step. Running it tells you how much unmanaged data is sitting in your vector store, the shadow data that a query could surface today with no policy in the way.
How it works and where it runs
Registration reads vector identifiers from the connector and creates managed resource records for the ones that are missing. It does not move or copy your vectors; the data stays in your store. It is available for any connector that exposes a listing capability, which today includes pgvector, Supabase, Neon, Pinecone, Qdrant, Azure AI Search, and Vertex AI Search.
Once registered, the new resources are ready for classification. Run classification suggestions to propose labels for review, then activate policies against them. The result is a corpus where nothing is ungoverned by default.
Frequently asked questions
Does retroactive registration move or re-embed my data?
No. It only reads vector identifiers and creates managed resource records in Gateco. Your vectors and their embeddings stay exactly where they are in your store.
Which connectors support it?
Any connector with a listing capability: pgvector, Supabase, Neon, Pinecone, Qdrant, Azure AI Search, and Vertex AI Search. Connectors without a listing API require resources to be registered through ingestion or the API.
What happens if the collection is empty or missing?
Registration handles an empty or not-yet-created collection gracefully by returning no resources, rather than erroring, so it is safe to run on a brand-new connector.