Abstract
This article describes a client-side system that discovers artificial-intelligence services available on the same device, on a local network, or through a configured remote connection, and selects a suitable service separately for each request. Selection considers required functions, input sensitivity, response delay, service load, operating cost, network condition, and the client device’s battery and thermal state. A local classifier applies mandatory handling rules before any request leaves the device. Services that fail a functional or data-location constraint are removed; the remaining services are ranked, and ordered alternatives are retained for controlled failover. Conversation state is stored in a service-neutral representation so that processing can continue when a service appears, disappears, changes address, or becomes unsuitable. The described architecture specifies discovery records, capability descriptors, registry timing, selection equations, request identifiers, context transfer, and failure-state behavior. The resulting application adapts at run time without manual endpoint changes while retaining control of request placement and session continuity.
Creative Commons License
This work is licensed under a Creative Commons Attribution-Share Alike 4.0 License.
Recommended Citation
INC, HP, "Adaptive Discovery and Policy-Based Routing of Local and Network AI Services", Technical Disclosure Commons, (September 07, 2026)
https://www.tdcommons.org/dpubs_series/11614