qualia_core_db/inference/runtime/scheduler/mod.rs
1//! Bounded request scheduling for paged native inference.
2//!
3//! Construction is cold and sizes every request table. Admission, prefix attachment, runnable
4//! collection, page COW, cancellation, and completion are zero-growth hot operations.
5
6mod batch;
7mod request_table;
8
9pub use batch::{
10 RaggedBackendError, RaggedBatchItem, RaggedBatchOutput, RaggedBatchReceipt, RaggedDecodeBackend,
11};
12pub use request_table::{
13 Admission, DecodeRoundError, RequestScheduler, RequestState, RequestView, SchedulerError,
14};
15
16#[cfg(test)]
17mod tests;