v0.2.1

Production Stack v0.2.1

Container image

ghcr.io/kaito-project/gpu-node-mocker:0.2.1

Helm charts

Add the chart repository (once):

helm repo add production-stack https://kaito-project.github.io/production-stack/charts/kaito-project
helm repo update production-stack

The following charts are published from this release (versions taken
from each chart's Chart.yaml at this tag):

• production-stack/gpu-node-mocker
• production-stack/modeldeployment
• production-stack/modelharness

See README.md for installation steps.

---

What's Changed

• feat(charts): add gateway error-mapping, ownership labels, and namespce management by @rambohe-ch in #99
• feat(charts): consolidate whole-path outage local_reply with component-first attribution by @rambohe-ch in #100
• release: bump charts to 0.2.1 by @rambohe-ch in #101

Full Changelog: v0.2.0...v0.2.1

v0.2.0

Production Stack v0.2.0

Container image

ghcr.io/kaito-project/gpu-node-mocker:0.2.0

Helm charts

Add the chart repository (once):

helm repo add production-stack https://kaito-project.github.io/production-stack/charts/kaito-project
helm repo update production-stack

The following charts are published from this release (versions taken
from each chart's Chart.yaml at this tag):

• production-stack/gpu-node-mocker
• production-stack/modeldeployment
• production-stack/modelharness

See README.md for installation steps.

---

What's Changed

• fix release step by @rambohe-ch in #54
• test(e2e): probe EPP pod for netpol deny assertions by @tnsimon in #53
• e2e: add provider switch (upstream/azure), bump keda-kaito-scaler to v0.5.1, instrument timings by @rambohe-ch in #50
• fix nightly e2e workflow by @rambohe-ch in #48
• chore: Surface error when docker or podman is not installed by @techworldhello in #56
• chore: add workflow to push helm charts to mcr by @t0rr3sp3dr0 in #63
• Replace model-not-found Service with Envoy direct_response EnvoyFilter by @rambohe-ch in #60
• feat: separate image preparation from cluster setup for accurate E2E timing by @rambohe-ch in #61
• feat: Garbage collector for GPU mocker by @techworldhello in #58
• feat: upgrade readme according to the latest helm chart by @rambohe-ch in #64
• Refactor E2E install around the productionstack umbrella chart by @rambohe-ch in #67
• Generate E2E coverage report by @techworldhello in #73
• Improve networkpolicy harness by @tnsimon in #57
• Pin BBR and keda-kaito-scaler versions in productionstack chart by @rambohe-ch in #69
• Make nightly env setup consistent to e2e by @techworldhello in #75
• test(e2e/netpol): dump canary pod sidecar state on enforcement precheck failure by @tnsimon in #77
• feat(productionstack): add llm-gateway-apikey as OCI Helm dependency by @tnsimon in #82
• Proposal: add End-to-End Error Handling Across Cluster, Modelharness, and Modeldeployment Levels by @rambohe-ch in #76
• fix label filters for report data by @techworldhello in #81
• fix(modelharness): scope NetworkPolicy selector to production-stack pods (#83) by @tnsimon in #84
• docs: recommend Cilium dataplane; update NetworkPolicy → CiliumNetworkPolicy references by @tnsimon in #85
• chore: bump llm-gateway-apikey to 0.0.10-alpha by @tnsimon in #93
• add highly-available requirements for llm-gateway-auth and bbr by @rambohe-ch in #86
• feat(bbr): harden body-based-router for HA and add e2e coverage (#89) by @rambohe-ch in #95
• chore: bump llm-gateway-apikey to 0.0.11-alpha; per-namespace ext_authz EnvoyFilter by @tnsimon in #97
• feat(charts/productionstack): fail-closed cluster filters + unified outage local_reply by @rambohe-ch in #94
• release: bump charts to 0.2.0 and publish productionstack by @rambohe-ch in #98

New Contributors

• @t0rr3sp3dr0 made their first contribution in #63

Full Changelog: v0.1.0...v0.2.0

v0.1.0

Production Stack v0.1.0

Container image

ghcr.io/kaito-project/gpu-node-mocker:0.1.0

Helm charts

Add the chart repository (once):

helm repo add production-stack https://kaito-project.github.io/production-stack/charts/kaito-project
helm repo update production-stack

The following charts are published from this release (versions taken
from each chart's Chart.yaml at this tag):

• production-stack/gpu-node-mocker
• production-stack/modeldeployment
• production-stack/modelharness

See README.md for installation steps.

---

What's Changed

• feat: add common ci and makefile snippets by @rambohe-ch in #2
• Create GPU mocker by @techworldhello in #1
• feat: add production stack arch contents by @rambohe-ch in #3
• Add GPU node mocker framework by @techworldhello in #4
• Fix rbac so that gpu node mocker can run by @techworldhello in #6
• feat: use llm-d-inference-sim as llm mocker by @rambohe-ch in #5
• feat: support verify boilerplate by @rambohe-ch in #8
• feat: add e2e tests framework by @rambohe-ch in #7
• fix: the error to extract model name from original pod by @rambohe-ch in #14
• E2E test setup for AKS cluster by @techworldhello in #16
• Add remaining repo setup by @techworldhello in #17
• Add local e2e setup by @techworldhello in #19
• feat: extract inferenceset part from e2e workflow to e2e cases by @rambohe-ch in #21
• feat: collect all versions of components by @rambohe-ch in #22
• Add gpu mocker tests by @techworldhello in #20
• feat: add keda and keda-kaito-scaler installation by @rambohe-ch in #23
• Add model routing e2e tests by @techworldhello in #26
• chore: bump keda-kaito-scaler to version v0.4.0 by @rambohe-ch in #28
• Add prefix cache aware routing tests by @techworldhello in #27
• fix: upgrade readme file by @rambohe-ch in #30
• feat: use CONTAINER_TOOL variable to support docker and podman by @tnsimon in #31
• fix: create shadow pods in model namespace instead of kaito-shadow by @tnsimon in #33
• fix: simplify e2e-up with username-based naming by @tnsimon in #34
• feat: add a modeldeployment helm chart by @rambohe-ch in #36
• feat: improve e2e tests by using modeldeployment helm chart by @rambohe-ch in #37
• feat: improve e2e tests by @rambohe-ch in #39
• feat: add e2e api key test by @tnsimon in #38
• feat: extract base setup workflow and improve install component in parallel by @rambohe-ch in #41
• feat: upgrade producation-stack arch and readme by @rambohe-ch in #43
• feat: add benchmark workflow by @rambohe-ch in #42
• fix the names for gateway host and steps by @rambohe-ch in #44
• fix: update llm-d-inference-sim in arch by @rambohe-ch in #45
• feat: add network policy e2e by @tnsimon in #25
• feat: add modelharness helm chart by @rambohe-ch in #46
• feat: add scaling test cases by @rambohe-ch in #40
• feat: upgrade arch with modelharness and modeldeployment by @rambohe-ch in #47
• chore: bump llm-gateway-auth to 0.0.7-alpha by @tnsimon in #49
• feat: add release workflow for production stack by @rambohe-ch in #52

New Contributors

• @techworldhello made their first contribution in #1
• @tnsimon made their first contribution in #31

Full Changelog: https://github.com/kaito-project/production-stack/commits/v0.1.0