InfinityAI and LLM Providers
Embeddings at Indexing Speed, on Your Own Servers
Infinity is a high-throughput inference server built for text embeddings. Connected to AI Intelligence Hub, it serves the embedding model behind retrieval on GPU servers you own. It is the choice for libraries where indexing volume is high enough that a general-purpose server would be the bottleneck. The text of transcripts, scanned pages, visual descriptions and documents becomes vectors on your hardware. The content itself stays in Nexus, and the language model that writes answers is configured separately.
What you can do together
- Index a large backlog by meaning in AI Intelligence Hub, with the embedding server sized for the volume rather than the volume waiting on the server.
- Keep every vector on your own hardware, connected or air-gapped.
- Pair Infinity with vLLM or Ollama so nothing about a question leaves your network.
- Get answers cited to the page or the moment, and bounded by the asking user's permissions.
How it connects
AI Intelligence Hub calls the Infinity server over its REST API on your network, at the address set in configuration. The embedding provider is configuration, so pointing the platform at Infinity rather than a hosted endpoint changes where embeddings are computed and nothing else.
At indexing time, the text of every stream goes to Infinity and vectors return to the index: transcripts, OCR text, visual descriptions and document text. Media does not go anywhere; the embedding call carries text only. At question time, the asking user's question is embedded the same way, and retrieval fuses that vector match with keyword matching over content the user is permitted to open.
Answers come from a separately configured language model. For a self-hosted deployment that is a vLLM or Ollama server, so retrieval and generation both stay inside the boundary.
InfinityAI and LLM Providers
A scenario
- SetupIT stands up Infinity on GPU servers in the court's data center and sets it as the embedding provider in AI Intelligence Hub. The Records agent's language model points at a vLLM server beside it.
- Weeks of scanningEach page is read by the platform's local OCR as it arrives, and its text is embedded through Infinity. The backlog indexes at the pace of the court's own servers. The scans stay in Nexus.
- 09:30A records clerk asks, "Which filings between 2009 and 2014 name this contractor?" The agent retrieves the matching pages she is permitted to open and sends them to vLLM with the question.
- Seconds laterThe answer lists the filings, each cited to a page. She opens a card and the scanned page opens.
- Same dayHearing recordings transcribed by the local Whisper model are embedded through the same server. A judge's clerk asks where a ruling on a motion was given, and the answer cites the moment in the recording.
- ThroughoutText moved between the platform, Infinity and vLLM on the court's network and nowhere else.
What stays where
Infinity runs on your servers
You choose the embedding model, load it and size the GPUs behind it for the indexing volume.
AI Intelligence Hub sends text across your own network
Text to embed goes to Infinity and vectors come back to the index. Media never leaves Nexus.
Nothing is replaced
The embedding provider is a configuration setting, and the agents, workflows and permissions around it stay as they are.
Where processing runs
On your servers, on-premises or in your own cloud subscription, connected or disconnected. Licensing verification is the platform's one outbound call, independent of AI processing, and it is disabled for a disconnected deployment.
Products and solutions
Next step
See it on your own Infinity 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