Want semantic search without operating a separate vector stack: no extra database, no standalone embedding service outside the platform.
Platform · Embeddings
Embeddings that make your knowledge findable
Searching thousands of documents takes more than keywords. Numaga turns your documents into embeddings (a numeric representation of meaning) so the knowledge base finds what you mean, even when the words differ. Through the same gateway as all your other AI traffic, so with the same control.
Who it’s for
Want people to find what is meant, including in large, Dutch-language document collections where keywords fall short.
Don’t want embedding cost as an invisible extra: reported separately, per model, apart from chat usage.
The problem
Embeddings are usually invisible, technically and financially.
Semantic search requires every document to be converted into embeddings. In most AI tools that happens out of sight: a separate service, a separate bill, and usage that never lands in your reporting. Large document collections make that painful: tens of thousands of documents mean tens of thousands of calls.
In Numaga, embeddings run through the same gateway as the rest of your AI traffic. They fall under the same management, storage sits in the platform’s managed database, and usage is reported separately, not mixed with chat, because the volumes and cost are fundamentally different.
How it works
From document to findable answer.
Automatic with every knowledge-base sync, without separate vector infrastructure.
Documents become embeddings
On every knowledge-base sync and every upload, the platform converts the content into embeddings via an embedding model on the gateway.
Stored inside the platform
The vectors live in your environment’s managed database: single-tenant, in the Netherlands, with no separate vector stack to maintain.
Search by meaning and exact terms
Questions are answered using a hybrid approach: semantic search combined with exact-term matching, which is important for names, codes and domain jargon.
Usage reported separately
Embedding traffic has its own overview: which models, how many calls and what it costs. It sits apart from chat usage, so your reporting stays clean.
In practice
Proven where data is most sensitive.
Numaga is running a pilot with Ontdek Zorg, a home-care organisation in Gelderland. Healthcare, with special-category personal data under the GDPR, is where responsible AI is hardest. That is exactly where we put the control plane to the test.
- Home care in Gelderland
- Special-category data (GDPR Art. 9)
- Policy and redaction on every prompt
Make your documents findable by meaning.
We show how embeddings work in your environment, from knowledge-base sync to the cost overview.
Request demo access