kserve/kserve v0.18.0-rc1 on GitHub

What's Changed

build(make): add optional overrides include to Makefile by @bartoszmajsak in #5164
fix: skip _folder to package helm chart by @Jooho in #5186
refactor(e2e/llmisvc): flexible endpoint testing by @bartoszmajsak in #5093
refactor: change kserve chart name to kserve-resources by @Jooho in #5190
feat(release): add draft release automation and improve release process by @Jooho in #5191
feat: 5060 CSV and Parquet marshallers by @cjohannsen-cloudera in #5115
feat(ci): conditionally trigger helm or kustomize e2e tests based on … by @Jooho in #5193
fix(llmisvc): resolve LWS leader address for ZMQ and increase /dev/shm by @bartoszmajsak in #5155
fix(utils): prevent overwriting existing SecurityContext when configu… by @fyuan1316 in #5167
chore: add MAINTAINERS.md for LFX Insights by @terrytangyuan in #5195
feat(logger): log occurrence and record time in CloudEvents by @samuelkim7 in #5139
fix(llmisvc): handle missing LWS CRD gracefully by @bartoszmajsak in #5201
feat: include GIE CRDs as part of Kserve bundle by @pierDipi in #5214
fix: add helm chart version to kserve-runtime-configs by @Jooho in #5209
modelcache: add Namespace Scoped ModelCache by @VedantMahabaleshwarkar in #4887
release: prepare release v0.17.0-rc1 by @Jooho in #5222
fix(llmisvc): add validations for unsupported scaling configurations by @vivekk16 in #5212
fix(make): increase test timeout from 20m to 30m by @bartoszmajsak in #5165
fix: improve cherry-pick workflow with validation and PR fetch by @Jooho in #5233
chore: bump kustomize from v5.5.0 to v5.8.0 by @bartoszmajsak in #5218
feat(llmisvc): add build hooks for distribution-specific logic by @bartoszmajsak in #5217
test(llmisvc): unit tests for versioned config resolution by @bartoszmajsak in #5182
fix: resolve go test -cover failure by @bartoszmajsak in #5178
fix(build): fall back to existing version when URL verification fails by @bartoszmajsak in #5180
feat: allow configuring GatewayClass for llmisvc e2e tests by @pierDipi in #5238
feat: upgrade llm-d to 0.6 by @pierDipi in #5121
fix(llmisvc): add default for MinReplicas & make MaxReplicas required int32 by @vivekk16 in #5237
feat(llmisvc): add --disable-uvicorn-access-log to vllm configs by @pierDipi in #5154
feat: merge target branch in each GH action job by @pierDipi in #5232
feat(llmisvc): implement autoscaling for WVA and KEDA/HPA integration by @vivekk16 in #5194
feat: add storage migration for LLMinferenceService APIs by @pierDipi in #5149
fix(llmisvc): add missing os.Exit(1) after controller setup failure by @vivekk16 in #5242
fix: change wrong crd kustomization update by @spolti in #5234
fix: to install specific kserve version with kustomize script by @Jooho in #5247
test(llmisvc): add envtest integration tests for scaling reconciliation by @vivekk16 in #5246
release: prepare release v0.17.0 by @Jooho in #5248
feat(llmisvc): evaluate both InferencePool readiness by @bartoszmajsak in #5202
fix(ci): configure git identity for merge commits in CI workflows by @pierDipi in #5251
feat(llmisvc): update LLMInferenceServiceConfig and add TLS flag by @pierDipi in #5249
chore: migrate Python formatter from black to ruff format by @jlost in #5000
fix(ci): add readiness to seaweedfs deployment by @pierDipi in #5267
feat: replace vllm-cpu image with upstream built image by @pierDipi in #5264
fix(e2e): retry predict on transient gateway errors by @pierDipi in #5266
feat: enable scheduler TLS cert reload by @pierDipi in #5260
feat: adding version overrides helm chart dependencies by @andresllh in #5268
fix(ci): restrict workflow re-run to repo collaborators by @bartoszmajsak in #5280
fix: prevent double-build when pushing llmisvc-controller image by @israel-hdez in #5207
chore(ci): align action versions across workflows by @bartoszmajsak in #5277
fix(ci): correct self-referencing path in llmisvc docker workflow by @bartoszmajsak in #5275
fix(ci): add missing disk space cleanup to docker push jobs by @bartoszmajsak in #5278
chore: support for merge queue by @israel-hdez in #5253
ci: optimize Go Dockerfiles for faster CI builds by @bartoszmajsak in #5273
modelcache: run namespace-scoped download jobs in jobNamespace by @VedantMahabaleshwarkar in #5262
chore(ci): remove redundant Go dependency download step by @bartoszmajsak in #5279
ci: improve GHA runner infra (disk cleanup, minikube addons) by @bartoszmajsak in #5281
feat(llmisvc): add extra_headers to wait_for_model_response by @jlost in #5215
feat(localmodel): add build hooks for distribution-specific logic by @VedantMahabaleshwarkar in #5263
[WIP] feat: detect Gateway for auto-migration to v1 InferencePool by @pierDipi in #5041
fix: include 0.17.0 installation scripts by @Jooho in #5256
fix: replace MD5 with SHA-256 in e2e fixtures by @maskarb in #5271
fix: "[WIP] feat: detect Gateway for auto-migration to v1 InferencePool" by @pierDipi in #5316
fix(llmisvc): migrate tokenProcessorConfig and blockSize by @pierDipi in #5305
ci: improvements for setup-uv and build-graph image by @bartoszmajsak in #5220
build: optimize Python Dockerfile caching and build context by @bartoszmajsak in #5274
chore: bump mlserver seldon image to latest 1.7.1 by @seBasKov in #4573
fix(llmisvc): autoscaling config logging by @Gregory-Pereira in #5304
fix(llmisvc): remove unreachable RefsInvalid early-return guards by @bartoszmajsak in #5288
chore(hack): remove legacy quick install scripts by @Jooho in #5289
refactor(test): migrate pod webhook suite to centralized envtest setup by @bartoszmajsak in #5179
chore: fix CVE-2026-32597 PyJWT crit header validation by @spolti in #5283
feat: update .gitattributes with all generated files by @pierDipi in #5258
fix(deps): bump go.opentelemetry.io/otel/sdk to v1.40.0 by @SebastienSyd in #5199
chore(config): clean up trailing whitespaces by @juhyeon-cha in #5300
fix(llmisvc): remove cert-hash restart annotation from scheduler by @bartoszmajsak in #5308
fix(llmisvc): defer storage migration until webhooks are serving by @bartoszmajsak in #5286
fix(controllers): gracefully handle resource deletion by @bartoszmajsak in #5252
fix(build): pin setup-envtest by @bartoszmajsak in #5323
chore: pin standalone kustomize and decouple from kubectl by @jlost in #5322
ci: reduce Go image build times by @bartoszmajsak in #5287
refactor(llmisvc): simplify EPPServiceName guard logic by @bartoszmajsak in #5107
feat(llmisvc): propagate imagePullSecrets from default SA by @pierDipi in #5324
test: centralizes logger setup by @bartoszmajsak in #5042
fix: merge automation by @pierDipi in #5328
ci: add pinact to pin GitHub Actions to SHA by @pierDipi in #5320
fix(test): ensures GinkgoRecover is used correctly by @bartoszmajsak in #5171
fix(ci): pins github actions in license check by @bartoszmajsak in #5329
fix(ci): 'commit not found' in go-security-scan upload-sarif by @Jooho in #5336
feat(llmisvc): upgrade WVA dependency from v0.5.1 to v0.6.0-rc3 by @vivekk16 in #5333
docs: add e2e guide for running GPT-OSS-20B with LLMInferenceService by @sbekkerm in #5150
fix: 5337: upgrade vLLM to 0.17.1 by @cjohannsen-cloudera in #5338
feat: add offset_mapping output for HuggingFace token classification by @Yuuuuuu0319 in #5244
fix(llmisvc): always set status.URL with a discovered address by @pierDipi in #5339
feat(ingress): add DisableHTTPRouteTimeout config flag for GKE Gateway by @sophieliu15 in #5313
feat: add /v1/responses HTTPRoute for OpenAI Responses API support by @RishabhSaini in #5291
chore(llmisvc): bump LLMISvc dependency versions by @vivekk16 in #5344
feat: block PYTHONPATH env var in ISVC and ServingRuntime webhooks by @Jooho in #5340
fix(llmisvc): upgrading to LLM-D v0.6.0 for llmisvc by @Gregory-Pereira in #5346
feat(llmisvc): add platform hooks for networking and manager cfg by @bartoszmajsak in #5240
feat(llmisvc): enable LWS as autoscaling target for multi-node workloads by @vivekk16 in #5356
fix: hardcode Status.Address scheme to http in raw deployment by @jlost in #5363
chore: remove coverage badge by @israel-hdez in #5298
feat(multinode): add groundwork for multinode support without ray by @Jooho in #5366
fix: upgrade vLLM to 0.19.0 by @cjohannsen-cloudera in #5367
fix: enforce PSS restricted profile on LLMInferenceService default template by @alokdangre in #5302
fix: recover multi download flag in default CSC by @israel-hdez in #5368
fix(agent): mirror user-defined httpGet path onto agent readiness probe by @reyshazni in #5345
fix(llmisvc): shorten webhook name to comply with 63-char label limit by @spolti in #5381
fix: CVE-2026-33186 gRPC authorization bypass by @sivanantha321 in #5342
docs: fix typo in OpenShift guide by @iranzo in #5197
revert(llmisvc): restore original webhook names, reverts #5381 by @bartoszmajsak in #5398
fix: intermittent 403 Forbidden errors during S3 model download by @agoston in #5393
feat(llmisvc): add SectionName support to Gateway refs by @bartoszmajsak in #5410
build: add GOTAGS build arg to manager and agent Dockerfiles by @bartoszmajsak in #5403
refactor(test): migrate credentials suite to NewEnvTest by @bartoszmajsak in #5397
fix: address CVE-2026-30922: pyasn1 Vulnerable to DoS by @spolti in #5404
feat(release): add Copilot CLI release automation by @Jooho in #5419
release: prepare release v0.18.0-rc0 by @Jooho in #5422
release: prepare release v0.18.0-rc1 by @Jooho in #5435

New Contributors

@fyuan1316 made their first contribution in #5167
@maskarb made their first contribution in #5271
@seBasKov made their first contribution in #4573
@Gregory-Pereira made their first contribution in #5304
@SebastienSyd made their first contribution in #5199
@Yuuuuuu0319 made their first contribution in #5244
@sophieliu15 made their first contribution in #5313
@RishabhSaini made their first contribution in #5291
@alokdangre made their first contribution in #5302
@reyshazni made their first contribution in #5345
@iranzo made their first contribution in #5197

Full Changelog: v0.17.0...v0.18.0-rc1