1pub use spec_stream_core as core;
18
19#[cfg(feature = "cache")]
20pub use spec_stream_cache as cache;
21
22#[cfg(feature = "engine")]
23pub use spec_stream_engine as engine;
24
25pub mod prelude {
27 pub use spec_stream_core::{
28 ByteLevelTokenizer, DraftProposal, Logits, SamplingParams, SpeculativeMetrics,
29 SpeculativeTree, SpeculativeVerifier, TokenCandidate, TokenId, Tokenizer,
30 TreeVerificationResult, VerificationResult, VerificationStrategy,
31 };
32
33 #[cfg(feature = "cache")]
34 pub use spec_stream_cache::{
35 BlockAllocator, BlockTable, EvictionPolicy, LruEvictionPolicy, PagedKvCacheManager,
36 PhysicalBlock, PhysicalBlockId, PrefixCacheSnapshot, PrefixSnapshotEntry,
37 };
38
39 #[cfg(feature = "engine")]
40 pub use spec_stream_engine::{
41 ChunkedPrefillConfig, ChunkedPrefillScheduler, ContinuousBatchScheduler, GenerationEvent,
42 GgufMetadata, LinearProjectionModel, ModelEngine, PromptLookupConfig, PromptLookupEngine,
43 QuantType, QuantizedLinearModel, QueuedRequest, SafeTensorsMetadata, SchedulerStats,
44 SpeculativeConfig, SpeculativePipeline, SpeculativeStream, SyntheticModel,
45 TreeSpeculativeConfig, TreeSpeculativePipeline,
46 };
47}