Runtime selection
Separate application-facing intent contracts from the model selected for execution.
Platform
Models + router
Change model strategy without changing product code.
Configure provider credentials and runtime preferences behind the intent contract. Route requests according to your operating needs while the application keeps calling the same capability.
Runtime routing
Ready
ProvidersModelsFallbacksHealth
Live configuration
primary preferred-model
fallback_1 secondary-model
fallback_2 recovery-model
fallback_1 secondary-model
fallback_2 recovery-model
ContractUnchanged
RoutingConfigured
Provider keysTenant-owned
Contract-stableModel changes stay behind the intent API.
Provider-awareConfigure credentials and preferences centrally.
Failure-consciousPlan fallback behavior before an outage.
Platform capabilities
Everything needed to operate this layer.
Build the capability once, keep its operating context visible, and improve it without spreading AI infrastructure across your application.
Fallback configuration
Define recovery paths so provider failure does not require emergency application changes.
Provider credentials
Manage tenant-owned provider access within the platform configuration boundary.
Model portability
Evaluate alternate models without rewriting every consuming application service.
Trace attribution
See which provider and model handled a request in its execution record.
Cost context
Connect model selection to the cost and usage data produced by real requests.
Operating workflow
From configuration to production evidence.
The interface follows the work in the order teams actually perform it.
01
Connect providers
Add the credentials your tenant controls.
02
Set preferences
Choose primary and fallback runtime options.
03
Test behavior
Exercise normal and failure scenarios.
04
Observe routing
Review model attribution in request traces.
What improves
A cleaner operating model for production AI.
Move complexity into a visible platform boundary while keeping the product integration direct.
01
Reduced provider coupling
Product integrations depend on intent contracts instead of provider-specific request shapes.
02
Planned resilience
Fallback choices are reviewed before an incident rather than improvised during one.
03
Better model decisions
Routing choices can be evaluated against quality, latency, and cost evidence.
Questions
Before you build.
Practical answers about how this part of LiyaEngine fits into the platform.
Does my endpoint change when I change models?+
No. The application continues to call the same domain and intent endpoint.
Can tenants use their own provider credentials?+
Yes. Provider access is configured for the tenant rather than embedded in application code.
Can I see which model handled a request?+
Yes. Provider and model attribution are included in execution visibility.
Build the capability. Keep the operational context.
Start with a tenant-owned domain and expose the first intent through a stable, governed API.