lib.rs 1.9 KB
Newer Older
1
2
3
4
5
6
7
8
// SPDX-FileCopyrightText: Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
// SPDX-License-Identifier: Apache-2.0

//! KV Router - Radix tree data structures for LLM KV cache routing.
//!
//! This crate provides the core radix tree implementation and protocols for
//! efficient KV cache lookup and routing in distributed LLM inference systems.

9
pub mod event_sink;
10
11
pub mod indexer;
pub mod protocols;
12
13
pub mod scheduling;
pub mod sequences;
14
pub mod zmq_wire;
15

16
17
18
19
20
21
22
23
// Backward-compat re-exports: old top-level module paths still work
pub use indexer::concurrent_radix_tree;
#[cfg(feature = "bench")]
pub use indexer::naive as naive_indexers;
pub use indexer::positional as nested_map;
pub use indexer::pruning as approx;
pub use indexer::radix_tree;

24
25
26
27
28
pub use scheduling::config;
pub use scheduling::queue;
pub use scheduling::selector;
pub use sequences::multi_worker as multi_worker_sequence;
pub use sequences::single as sequence;
29

30
31
#[cfg(any(test, feature = "bench"))]
pub mod test_utils;
Yan Ru Pei's avatar
Yan Ru Pei committed
32

33
// Re-export key types for convenience
34
pub use self::multi_worker_sequence::{
35
36
37
    ActiveSequencesMultiWorker, SequenceError, SequencePublisher, SequenceRequest,
    SequenceSubscriber,
};
38
39
40
pub use self::sequence::{ActiveSequences, RequestId};
pub use concurrent_radix_tree::ConcurrentRadixTree;
pub use config::{KvRouterConfig, RouterConfigOverride};
41
pub use event_sink::EventSink;
42
pub use indexer::{MaybeError, SyncIndexer, ThreadPoolIndexer};
43
44
#[cfg(feature = "bench")]
pub use naive_indexers::{InvertedIndex, NaiveNestedMap};
Yan Ru Pei's avatar
Yan Ru Pei committed
45
pub use nested_map::PositionalIndexer;
46
pub use protocols::{
47
    KvCacheEventError, LocalBlockHash, OverlapScores, RouterEvent, WorkerConfigLike, WorkerId,
48
49
    compute_block_hash_for_seq,
};
50
pub use queue::SchedulerQueue;
51
pub use radix_tree::RadixTree;
52
53
pub use scheduling::{KvSchedulerError, PotentialLoad, SchedulingRequest, SchedulingResponse};
pub use selector::{DefaultWorkerSelector, WorkerSelector};