Azure OpenAI ServiceAI and LLM Providers
Embeddings That Never Leave Your Azure Tenancy
Azure OpenAI Service offers the OpenAI models inside your own Azure subscription, in a region you choose, under that subscription's regional commitments. Connected to AI Intelligence Hub, it supplies the embedding models behind retrieval. The text of your transcripts, scanned documents and visual descriptions is turned into vectors inside your tenancy rather than at a public endpoint. The recordings and documents stay in Nexus under your access rules, and the language model that writes answers is chosen separately, per agent.
What you can do together
- Search by meaning across AI Intelligence Hub with every vector computed in your own Azure subscription and region.
- Pair those embeddings with the answer model of your choice per agent: a hosted provider for general material, a self-hosted server for what must stay on your network.
- Get answers grounded in retrieved passages and cited to the moment or the page, rather than generated from a model's training.
- Keep an agent inside a folder, category, tag set or attribute value, and inside what the asking user is entitled to open.
How it connects
You create an Azure OpenAI resource in your subscription and deploy an embedding model to it. AI Intelligence Hub calls that resource's REST API endpoint with the key you supply. The embedding provider is configuration, and this is the usual choice where an organization needs embedding calls to stay inside a tenancy and region it controls.
At indexing time, text goes to your Azure OpenAI endpoint and vectors return to the index. Every stream is embedded: transcripts, OCR text, visual descriptions and document text. Media does not go; the embedding call carries text only.
Answers are a separate setting. The model behind each agent is configured from the platform's language model providers, hosted or self-hosted. Retrieval always runs under the asking user's identity, with their access list applied before the vector search.
Azure OpenAI ServiceAI and LLM Providers
A scenario
- SetupIT deploys an embedding model to the bank's Azure OpenAI resource and enters the endpoint and key in AI Intelligence Hub. The Compliance agent gets a self-hosted language model on the bank's vLLM server, scoped to the recorded-calls and complaints folders.
- OvernightThe day's advisor calls are transcribed on ingest, the scanned complaint letters are read by OCR, and all of that text is embedded through Azure OpenAI. Vectors join the index; the audio and scans stay in Nexus.
- 08:45A compliance investigator asks, "Find every call last month where an advisor described a fund as guaranteed." Retrieval fuses keyword and vector matching over the calls she is permitted to hear.
- Seconds laterThe answer lists the calls, each cited to the moment the phrase was spoken. She opens a card and the call plays from that second.
- 09:00She asks whether any of those customers later complained. The agent retrieves the matching letters and answers with a citation to each page. Text went to the bank's own Azure region and to its own server; nothing left the tenancy.
What stays where
Azure OpenAI Service stays in your subscription
The resource, the region, the key and the quota are yours.
AI Intelligence Hub sends text to embed and gets vectors back
Media never goes to the endpoint. The index and the content sit in your deployment.
Nothing is replaced
The embedding provider is configuration; the agents, workflows and permissions around it do not change.
Where processing runs
Embedding runs in the Azure region your resource is bound to. Answer generation runs wherever the model you chose per agent runs: a hosted provider's service, or a vLLM or Ollama server on your own hardware. An air-gapped deployment cannot reach Azure and uses a self-hosted embedding provider instead.
Products and solutions
Next step
See it on your own Azure OpenAI Service instance.
We will show the connection made, the data moving and the output, then size it for your deployment.
Contact VIDIZMO
sales@vidizmo.ai
+1 571-969-2180
vidizmo.ai