| vllm-project/vllm-omni @ `18c3bd5d2d4323bc7adca1ac3c465c4a53bbfb6e` (Apache-2.0) | Engine-base abort/ack APIs, correlated waiter/router, orchestrator companion and logical cleanup, stage-pool partial outputs and queued release; selected upstream test source and test bootstrap | Extend [existing omni-validation](../src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/SKILL.md) with [three bounded mechanisms and pinned readset](../src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/abort-acknowledgment.md). Reuse the original `52194f24...` engine guide and catalog ownership; no new skill or parallel database. CODE / TEST-SOURCE only; actual device, connector, model, service and performance not_run. Full duplex/session behavior and native abort guarantees deferred. |
0 commit comments