All issues

Search results

fix(scheduler): keep tenant queue caps opt-in

Status: Open.

#48 In LLM360/smg;

· DavidBellamy opened on Aug 6, 2026

fix(admission): dispatch fair-share queues at engine-feedback capacity

Status: Open.

#45 In LLM360/smg;

· DavidBellamy opened on Aug 6, 2026

KV event gaps trap cache-aware workers in a replay reconnect loop

Status: Open.

#42 In LLM360/smg;

· DavidBellamy opened on Aug 5, 2026

feat(admission): add bounded prediction samples

Status: Open.

#35 In LLM360/smg;

· DavidBellamy opened on Aug 4, 2026

feature: adaptive model admission from predicted token work and engine pressure

Status: Open.

#30 In LLM360/smg;

· DavidBellamy opened on Aug 3, 2026

Make priority admission queue work-conserving and add Retry-After estimates

Status: Open.

#23 In LLM360/smg;

· DavidBellamy opened on Aug 2, 2026

Expose priority scheduler settings through Python router entrypoint

Status: Open.

#17 In LLM360/smg;

· DavidBellamy opened on Jul 31, 2026

[Feature]: make cache-aware routing engine-pressure aware

enhancementNew feature or requestNew feature or request

triage

Status: Open.

#11 In LLM360/smg;

· DavidBellamy opened on Jul 28, 2026