All pull requests

Pull requests

• fix(runtime): make the CallHome listen backlog configurable, default 4096

  fixsize/M
  #15270·tzulingk opened Sep 25, 2026Contributor·

  ·
  3

• fix(runtime): retry NATS connections during startup

  deployment::k8sdocumentationexternal-contributionfixsize/XXL
  #15269·surajssd opened Sep 25, 2026·

  ·
  1

  6

• docs(sglang): add EAGLE3 speculative decoding guide

  backend::sglangdocsdocumentationexternal-contributionsize/L
  #15268·livingshade opened Sep 25, 2026·

  ·
  8

• fix(runtime): recover hung etcd watches with keepalive

  external-contributionfixsize/Ltrusted-contributor
  #15266·michaelfeil opened Sep 24, 2026Contributor·

  ·
  1

  8

• feat(llm): add Mocker prefix-cache query/hit metrics

  external-contributionfeatsize/L
  #15264·patricklundquist opened Sep 24, 2026·

  ·
  11

• fix(runtime): detect hung etcd members with channel keepalive; never abandon discovery watches

  external-contributionfixsize/L
  #15263·devabhixda opened Sep 24, 2026·

  ·
  1

  4

• test(sglang): add EAGLE3 disagg KV-router e2e

  backend::sglangexternal-contributionsize/Ltest
  #15261·livingshade opened Sep 24, 2026·

  ·
  13

• feat(kv-router): preserve shared-cache event eligibility

  external-contributionfeatroutersize/XXL
  #15260·surajssd opened Sep 24, 2026·

  ·
  5

• fix(vllm): allow sparse KV totals in explicit prefill rows

  backend::vllmfixsize/XS
  #15259·PeaBrane opened Sep 24, 2026Contributor·

  ·
  1

• fix(kv-router): break request-lease observer ownership cycle

  external-contributionfixroutersize/L
  #15258·surajssd opened Sep 24, 2026·

  ·
  3

• fix(vllm): warm Inkling convolution caches with native prefills

  backend::vllmdocumentationfixsize/XXL
  #15257·Arsene12358 opened Sep 24, 2026Contributor·

  ·
  1

• chore(vllm): use explicit NIXL roles in runtime configurations

  backend::vllmchoredeployment::k8sdocumentationexternal-contributionplannersize/L
  #15256·junuxyz opened Sep 24, 2026·

  ·
  20

• perf(frontend): auto-size SGLang preprocess workers

  documentationexternal-contributionfrontendperfsize/L
  #15253·cmdy opened Sep 24, 2026Contributor·

  ·
  9

• fix(vllm): declare the ModelExpress startup weight version

  backend::vllmdocumentationexternal-contributionfixsize/L
  #15252·c2w-sea opened Sep 24, 2026·

  ·
  8

• test(sglang): add EAGLE3 speculative decoding e2e

  backend::sglangexternal-contributionsize/Ltest
  #15251·livingshade opened Sep 24, 2026·

  ·
  10

• fix(frontend): Flush buffered byte-fallback decoding

  external-contributionfixsize/L
  #15249·alexeldeib opened Sep 24, 2026Contributor·

  ·
  8

• fix(request-trace): emit request-end rows for Python chat processors

  documentationexternal-contributionfixsize/L
  #15248·fengyuli-dev opened Sep 24, 2026Contributor·

  ·
  7

• fix(frontend): reject continuation across reordered messages

  external-contributionfixfrontendsize/M
  #15247·alexeldeib opened Sep 24, 2026Contributor·

  ·
  4

• docs: move Enterprise artifacts to 1.5.0

  docsdocumentationsize/M
  #15245·dagil-nvidia opened Sep 24, 2026Collaborator·

  ·
  1

• fix: stop CODEOWNERS edits from requesting every owner

  fixsize/M
  #15244·dagil-nvidia opened Sep 24, 2026Collaborator·

  ·
  5

• test(sidecar): add isolated units beside production code

  actionsdocumentationsize/XXLtest
  #15243·JulienDarve opened Sep 24, 2026Contributor·

  ·

• build: compile NVTX ranges into runtime wheels via the header-only nvtx crate

  buildcontainersize/M
  #15242·peilii opened Sep 23, 2026Contributor·

  ·
  13

• fix(anthropic): open a fresh text block for text arriving after streamed tools

  external-contributionfixfrontendsize/Mtrusted-contributor
  #15241·michaelfeil opened Sep 23, 2026Contributor·

  ·
  8

• feat(replay): add conversation affinity to the existing adapter

  containerdocumentationfeatplannerroutersize/XXL
  #15240·Harrilee opened Sep 23, 2026Contributor·

  ·

• feat(mocker): support state-cache checkpoint reuse

  containerdocumentationfeatplannersize/XXL
  #15239·dreamtalen opened Sep 23, 2026Contributor·

  ·
  1