OpenAIAI and LLM Providers
GPT Models on Your Content, With Every Source Cited
OpenAI serves the GPT model family and embedding models over a hosted API. Connected to AI Intelligence Hub, OpenAI does two jobs. It embeds your content so retrieval works by meaning, and it writes the answer to a question from the passages retrieved for it, with a citation to each. The recordings and documents themselves stay in Nexus under your access rules. Which provider runs a request decides where that request is processed, and with OpenAI that is OpenAI's hosted service.
What you can do together
- Ask AI Intelligence Hub a question across call recordings, scanned letters and policy documents and get one answer with a citation card for each source.
- Find material by meaning rather than by exact words, because the embeddings index every stream: what was said, what was on screen, what a scene showed and what a document says.
- Run one agent on GPT for the shared knowledge base and another on a self-hosted model for the material that must stay inside the building.
- Attach a document or image in the conversation for analysis and keep asking follow-up questions in the same context.
- Build a workflow with a GPT node in it and a human approval step before the result is published.
How it connects
AI Intelligence Hub calls OpenAI's REST API with your organization's own account and key. The language model is configuration, set per deployment or per agent, so different agents can run different models. The embedding provider is configured separately and can be OpenAI as well.
Two things flow. At indexing time, the text of transcripts, OCR output, visual descriptions and documents goes to the embedding endpoint and vectors come back to the index. At question time, the question and the passages retrieved for the asking user go to the GPT model and the answer returns, cited to the passages it drew on.
Retrieval runs under the asking user's own identity, with their access list applied before the vector search. Media files never go to OpenAI, only text, and the library stays where it is.
OpenAIAI and LLM Providers
A scenario
- SetupIT enters the insurer's OpenAI key in AI Intelligence Hub and selects an OpenAI embedding model for the index. A GPT model becomes the model for the Claims Desk agent, scoped to the claims and policy folders.
- OvernightNew call recordings are transcribed on ingest, and their transcripts, with the day's scanned letters, are embedded through OpenAI. The vectors land in the index; the audio stays in Nexus.
- 08:30An adjuster asks, "What did the claimant say about when the water damage started, and what does the policy say about late reporting?" The agent retrieves the call passages and the policy clause she is permitted to see and sends them to GPT with the question.
- Seconds laterThe answer quotes the call, citing the moment in the recording, and the clause, citing the page. She opens the citation card and the call plays from that second.
- 08:40She attaches the contractor's estimate and asks whether the line items match what the claimant described. The conversation carries the context, and the answer cites both.
- LaterA trainee asks the same agent and gets an answer drawn only from the claims his role can open. The prompts and text went to OpenAI; no recording or scan left the insurer's deployment.
What stays where
OpenAI keeps serving the models
The account, the key, the usage and the choice of model per agent are yours.
AI Intelligence Hub sends text, not media
Text to embed goes out and vectors come back; a question and its retrieved passages go out and an answer comes back. Nothing else travels.
Nothing is replaced
Changing the model behind an agent is a configuration change, and the workflows behind it stay as they are.
Where processing runs
OpenAI processes prompts and embeddings in its hosted service. The platform runs as SaaS, in your own cloud subscription, on your own servers, or air-gapped. An air-gapped deployment cannot reach a hosted API and runs its models through Ollama or vLLM instead.
Products and solutions
Next step
See it on your own OpenAI 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