Expand description
Core type definitions for the Ferrum inference framework
This crate contains all the fundamental types, IDs, configurations, and error definitions that are shared across the entire Ferrum ecosystem. It is designed to be lightweight and dependency-free to avoid circular dependencies.
Re-exports§
pub use auto_config::*;pub use config::*;pub use devices::*;pub use errors::*;pub use execution_policy::*;pub use ids::*;pub use metrics::*;pub use models::*;pub use native_operator::*;pub use observability_profile::*;pub use process_memory::*;pub use reasoning::*;pub use requests::*;pub use resource_trace::*;pub use runtime_config::*;pub use sampling::*;
Modules§
- auto_
config - Startup auto-configuration and selector decision trace types.
- config
- Configuration types for Ferrum components
- devices
- Device and computation types
- errors
- Error types for Ferrum inference framework
- execution_
policy - Typed execution-policy and admission authority contracts.
- ids
- Identifier types for Ferrum entities
- metrics
- Metrics and observability types
- models
- Model-related types and configurations
- native_
operator - Native operator artifact manifest types.
- observability_
profile - Product-path observability profile event schema.
- process_
memory - reasoning
- Shared parsing for model outputs that carry
<think>reasoning blocks. - requests
- Request and response types for inference
- resource_
trace - Shared resource lifecycle event envelope for offline invariant checks.
- runtime_
config - Runtime configuration snapshot and small env parsing helpers.
- sampling
- Sampling and generation parameters