COMMITS
October 3, 2026
N
October 2, 2026
B
fix(multimodal): prevent ring buffer rewrap over live allocations (#15498)
Bojiang Li committed
J
test(sidecar): add vLLM CPU integration coverage (#15091)
JulienDarve committed
D
B
perf(cache): avoid scanning all keys during embedding admission (#15497)
Bojiang Li committed
Y
fix(kv-router): tighten slot tracker replica sync (#15578)
Yan Ru Pei committed
P
N
fix(multimodal): nvbug/6169700 cap the size of inline data: URLs (#14437)
Neelay Shah committed
N
refactor(request-trace): make payload capture a pass-through for non-streaming (#15397)
Neelay Shah committed
N
T
ci(deploy): read operator readiness from the vCluster Deployment (#15429)
Tushar Sharma committed
D
fix(benchmarks): OPS-8740 drop httpx backend attribution from the HTTP sweep harness (#14886)
Dmitry Tokarev committed
J
feat(mocker): publish native KV events from SGLang gRPC servers (#14744)
jthomson04 committed
J
ci(sidecar): run GPU E2E tests only post-merge and nightly (#15456)
JulienDarve committed
D
chore: bump trtllm to v1.3.0rc29 (#15418)
Dynamo Bot committed
N
fix(vllm): propagate FPM worker_id into snapshot-restored EngineCore (#12545)
Nianheng Nicole Wu committed
Y
perf(kv-router): rework worker recovery snapshot cache and relax dump limit (#15551)
Yan Ru Pei committed
C
fix(frontend): report dropped GLM-4.7 calls as length (#15184)
Chi McIsaac committed
D
feat(operator): add role-specific multinode PodTemplates (#14468)
Dr. Stefan Schimanski committed
D
docs: refresh community events
dynamo-ops committed
A
fix(vllm): accept Dynamo parser flags in sidecar (#15539)
Alec committed
G
fix(docs): let recipe cards without logos use full header width (#15546)
Glen Maddern committed
B
docs(sglang): add EAGLE3 speculative decoding guide (#15268)
Banruo Liu committed
B
test(sglang): add EAGLE3 disagg KV-router e2e (#15261)
Banruo Liu committed
E
fix(frontend): preserve colliding Responses tool namespaces (#15401)
Elizabeth Thomas committed
N
fix(frontend): pass checkpoint tokenizer to SGLang parsers (#15509)
neilkg committed
N
fix(frontend): classify SGLang chat template errors (#15511)
neilkg committed
C
fix(vllm): put omni video negative_prompt in prompt dict entry (fixes #15349) (#15402)
Charan Rathore committed
N
P
fix(vllm): use native Qwen3-ASR realtime support (#15481)
ptarasiewiczNV committed