mod config;
mod driver;
pub mod protocol;
mod token_emission;
pub use config::{
CacheRuntimeConfig, ContextRuntimeConfig, FlashAttentionMode, GenerateOptions, GpuLayerConfig,
KvCacheType, KvReuseMode, LogitBias, ModelLoadMode, ModelPlacementConfig,
MultimodalRuntimeConfig, NativeRuntimeConfig, ObservabilityRuntimeConfig,
ResidencyRuntimeConfig, ResolvedRuntimeLimits, RopeScaling, SamplerStage,
SamplingRuntimeConfig, SamplingRuntimeOverride, SchedulerRuntimeConfig, SplitMode,
DEFAULT_CONTEXT_KEY, DEFAULT_MAX_TOKENS,
};
pub use driver::{
ChatMessage, ChatRequest, ChatRole, EngineEmbeddingResponseFuture, EngineEmbeddingRun,
EngineEventReceiver, EngineLoad, EngineTextResponseFuture, EngineTextRun, EngineTokenBatches,
QueryOptions, QueryRequest, SippEngine,
};
pub(crate) use driver::{EngineAudioRun, EngineCancellation, ListenRequest, SpeakRequest};
pub use protocol::{
CacheSource, EmbedOptions, EmbedRequest, EmbeddingCapabilities, EmbeddingResult, EngineEvent,
EngineState, EngineStats, FinishReason, GenerationResult, ModelCapabilities, ModelClass,
PoolingType, RequestStats,
};
pub use token_emission::{TokenBatch, TokenEmissionStats};
#[cfg(test)]
#[path = "../tests/engine_tests.rs"]
mod engine_tests;