Multiple providers
Evaluate different hosted models against the same application task.
A model-flexible application layer, without false availability claims.
The direction is to keep product workflows stable while models and providers change. Compatibility depends on the real provider interface and must be verified rather than assumed.
Technical direction
Maturity is stated above. Research and planned work are not presented as publicly available platform features.
Evaluate different hosted models against the same application task.
Choose models using explicit capability, reliability and cost constraints.
Support OpenAI-compatible and, where appropriate, Anthropic-compatible request patterns after verification.
Keep domain logic outside provider-specific transport code.
Use measured task fit rather than a single global default.
Leave room for self-hosted or private inference when it becomes operationally justified.
Boundary
No credentials, private infrastructure addresses, routing configuration or operational secrets are exposed here.
Related architecture