Local + Cloud AI
Choose the model location. Keep the control path explicit.
A local model and a cloud model answer different operational needs. Kayrix is designed to make the provider configuration part of the controlled workflow rather than a choice hidden inside each application.
Local or on-premises
Evaluate a supported local model endpoint for use cases where deployment control matters. A local model still needs access controls, maintenance and an agreed data-handling configuration. Local does not automatically mean isolated, secure or legally compliant.
Cloud, under a defined configuration
Evaluate a supported cloud provider where its capabilities fit the task and the organization has approved the processing arrangement. Gateway controls address the requests Kayrix handles; they do not replace review of the provider’s terms, retention settings or responsibilities.
Integrations to scope in the private pilot
Ollama / local endpoints: candidate local inference path, subject to the approved integration and model configuration.
OpenAI: candidate cloud integration, subject to credentials, supported APIs and pilot validation.
Anthropic: candidate cloud integration, subject to credentials, supported APIs and pilot validation.
These names identify integration candidates, not commercial partnerships or endorsements. Availability and supported behavior are confirmed for each pilot; no model catalogue or universal compatibility is implied.
Additional providers: considered through future compatibility review. No availability date is promised.
Make the decision on operational grounds.
Describe the task, data sensitivity, required output and deployment constraints. Select a supported provider path and evaluate it against the agreed criteria. This is not a promise of autonomous optimal routing or equivalent results across providers.