--- title: "AI Model Cluster · AssistantGeneral" description: "Pool graphics cards and computers into one model cluster: local and remote nodes, topic tags, keyword routing, a memory budget, and helpers spread across them." url: "https://royalsoftworks.com/products/assistant-general/features/model-cluster/" source: "https://royalsoftworks.com" format: "markdown" note: "Markdown rendering of the HTML page at `url`. Same content, same canonical URL." --- Features · The AI itself # Several machines, one assistant. Pool graphics cards and computers into one model cluster: local and remote nodes, topic tags, keyword routing, a memory budget, and helpers spread across them. [Editions](https://royalsoftworks.com/products/assistant-general/editions/): Pro, Enterprise and Government. Not in Free. ## Model cluster, feature by feature Each comes with a small drawing of how it works. They are diagrams, not screenshots; numbers in them are examples. - ### Nodes on this computer Each node runs its own model file on its own port, with its own context size and graphics-card share. Leave the share blank to run on the processor. For your IT team: Per node: model file, optional vision projector, port, context tokens, GPU layers, a VRAM estimate. Diagram: This PC leads to Node 1 · GPU, Node 2 · GPU, Node 3 · CPU. - ### Nodes on other machines Add any model server by address, model name and key, and test it: the app reports whether it answered and how fast. Diagram: Add a remote model server, test it, and see whether it answered and how fast (42 ms is an example). - ### Routing by topic Tag nodes with what they are good at (code, maths, medical) and requests go to the best match. Diagram: A code question goes to Code node, not Maths node or Medical node. - ### Your own routing rules Send any request that mentions certain words to the node you choose. Diagram: Mentions “invoice” → Your rule → Finance node. - ### A graphics-memory budget Set how much graphics memory the cluster may use, and nodes take turns within it instead of crashing each other. Diagram: Example: two nodes using 8 GB and 6 GB inside a 16 GB graphics-memory budget. - ### Helpers across machines Let helpers run on the other nodes, so more than one part of a job is written at the same time. Diagram: Your request leads to Helper · PC 1, Helper · PC 2, Helper · PC 3. - ### Start and stop each node Every node switches on and off on its own, and can be disabled without being deleted. Checklist: Node 1 · on (yes), Node 2 · on (yes), Node 3 · off (no). ## Goes well with - [Local AI models](https://royalsoftworks.com/products/assistant-general/features/local-ai-models/) - [Helpers (subagents)](https://royalsoftworks.com/products/assistant-general/features/helpers/) - [Cloud AI providers](https://royalsoftworks.com/products/assistant-general/features/cloud-ai-providers/) [Previous Cloud AI providers](https://royalsoftworks.com/products/assistant-general/features/cloud-ai-providers/) [Next The agent pipeline](https://royalsoftworks.com/products/assistant-general/features/agent-pipeline/) [All 31 feature classes](https://royalsoftworks.com/products/assistant-general/features/)