1pub mod accelerator;
32pub mod api_response_cache;
33pub mod autotune;
34pub mod circuit_breaker;
35pub mod cluster;
36pub mod dashboard;
37pub mod discovery;
38pub mod health;
39pub mod host;
40pub mod kv_cache;
41pub mod load_balance;
42pub mod mdns;
43pub mod models;
44pub mod planning;
45pub mod protocol;
46pub mod ring;
47pub mod runtime;
48pub mod system_profile;
49pub mod watcher;
50#[doc(hidden)]
51pub mod xdp;
52
53pub use accelerator::ExecutionBackend;
55pub use api_response_cache::{ApiResponseCache, CacheStats};
56pub use autotune::AutoTuner;
57pub use circuit_breaker::{CircuitBreaker, CircuitBreakerConfig, CircuitState, JitteredBackoff};
58pub use cluster::{ClusterState, NodeMetrics, NodeStatus};
59pub use discovery::{broadcast_and_collect, UdpDiscoveryConfig, DEFAULT_DISCOVERY_PORT};
60pub use host::{
61 detect_local_node_resources, detect_runtime_profile, detect_runtime_profile_full,
62 detect_runtime_profile_with_mode, infer_compute_capability_from_name, is_apple_silicon,
63 AccelerationMode, GpuBackend, ProbeMode, RuntimeProfile,
64};
65pub use kv_cache::{KVCacheConfig, KVCacheEntry, LayerKvCache};
66pub use planning::{
67 assign_layers_sequentially, assign_layers_with_fault_tolerance_and_runtime,
68 assign_layers_with_runtime_profile, chunk_assignments_for_workers, select_quantization_mode,
69 LayerAssignment, LayerSpec, PlacementPlan, PlanningTuning, QuantizationMode,
70};
71pub use protocol::NodeResources;
72pub use ring::{RingConfig, SpscRingBuffer};
73pub use runtime::{
74 build_token_schedule, execute_pipeline, execute_pipeline_tcp_loopback,
75 execute_pipeline_tcp_loopback_with_config, BridgeAddr, BridgeListener, BridgeStream,
76 DeviceKind, ExecutionResult, PipelinePlan, StageExecutionStats, StagePlacement,
77 TcpTransportConfig, TokenStep, TransportKind,
78};
79pub use system_profile::{
80 CpuFeatures, CpuInfo, GpuInfo, MemoryInfo, NetworkInfo, NpuInfo, SystemProfile,
81};
82pub use watcher::{ProfileChange, SystemProfileWatcher};