[Router] Add load-aware prefill admission and bounded policy proposals (#37843)

Signed-off-by: Vincent Gao <vincentbo@linux.alibaba.com>
Co-authored-by: Kangyan Zhou <zky314343421@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Shangming Cai <csmthu@gmail.com>
This commit is contained in:
Vincent Gao
2026-09-05 11:53:26 +08:00
committed by GitHub
co-authored by Kangyan Zhou Claude Opus 4.8 Shangming Cai
parent d50e9a9756
commit ecd97de1fc
55 changed files with 6834 additions and 348 deletions
@@ -17,6 +17,7 @@ use sgl_kv_indexer::{
server_builder, GrpcPrefixIndex, InMemoryKvIndexerBackend, KvIndexerService, PrefixIndexConfig,
};
use sgl_router::discovery::{ModelId, WorkerId, WorkerMode, WorkerSpec};
use sgl_router::policies::engine_load::EngineLoadTable;
use sgl_router::policies::factory::build_registry;
use sgl_router::policies::kv_events::{compute_block_hashes, BlockSizeOracle, HashTree};
use sgl_router::policies::request_tokens_for;
@@ -94,6 +95,7 @@ async fn external_indexer_routes_to_the_cached_worker() {
Arc::new(HashTree::new()),
Arc::clone(&tokenizers),
Arc::clone(&oracle),
EngineLoadTable::new(),
)
.unwrap(),
);