Expand description
The assembled AI subsystem: the model registry, the provider clients that back it, the shared result cache, and the call log.
Built once from server configuration and threaded into the engine. Given a
model name referenced in SQL, AiRuntime::resolve returns everything the
inference operator needs to run it: the backend kind, a stable cache id, the
provider client, the provider-side model id, and any labels. The registry is
kept whole (it also backs the laminar.models catalog view), even for models
whose backend has no provider wired yet.
Structs§
- AiRuntime
- The assembled AI subsystem.
- Resolved
Model - Everything the inference operator needs to run one model.
Enums§
- AiRuntime
Error - Errors from resolving a model to a runnable backend.