llm-d is a Kubernetes-native high-performance distributed LLM inference framework

What's Changed

• Enhance dependency installation script by @ErikJiang in #366
• fix: Enable HTTP proxy support by removing sudo from curl commands in quickstart/install-deps.sh by @yankay in #370
• repo migration basic docs by @Gregory-Pereira in #372
• Fix e2e-validate permissions by @nerdalert in #374
• bump llm-d-x images to latest tags & improve ci by @nicole-lihui in #365

New Contributors

• @ErikJiang made their first contribution in #366
• @nicole-lihui made their first contribution in #365

Full Changelog: llm-d-1.0.22...llm-d-1.0.23

llm-d is a Kubernetes-native high-performance distributed LLM inference framework

What's Changed

• Fix Docs: Minikube gateway access instructions in quickstart guide by @yankay in #362
• Add probes to prefill and decode pods by @namasl in #351

New Contributors

• @yankay made their first contribution in #362
• @namasl made their first contribution in #351

Full Changelog: llm-d-1.0.21...llm-d-1.0.22

llm-d is a Kubernetes-native high-performance distributed LLM inference framework

What's Changed

• do not change user's default namespace by @sozercan in #336
• bugfix: set the namespace while creating the hf_token secret by @nerdalert in #340
• Fixes #347 by creating a download job on correct namespace by @maugustosilva in #348
• Fix prometheus metrics on OpenShift and simplify logic by @dagrayvid in #329
• Charts: Bumps Kgateway to v2.0.3 by @danehans in #353
• Istio bump to latest upstream alpha with GAIE features by @nerdalert in #355
• Add support for GKE Gateway API by @maci0 in #359

New Contributors

• @sozercan made their first contribution in #336
• @dagrayvid made their first contribution in #329
• @maci0 made their first contribution in #359

Full Changelog: llm-d-1.0.20...llm-d-1.0.21

llm-d is a Kubernetes-native high-performance distributed LLM inference framework

What's Changed

• [quickstart] Add option -j/--gateway to llmd-installer.sh by @maugustosilva in #316
• Fix: --download-model CLI parameter now correctly overrides template … by @yossiovadia in #320
• Add "release name" as cli parameter (-r/--release) for llmd-installer.sh by @maugustosilva in #326
• Add the ability to dispatch a PR or branch in ec2-e2e by @nerdalert in #325
• e2e lint fix by @nerdalert in #333
• fix: populate gateway.gatewayClassName helm value when the gateway type is not istio by @chewong in #327
• fix: correct --download-model to use HuggingFace URI instead of PVC by @yossiovadia in #328
• Detect minikube context for uninstall by @nerdalert in #278
• Add the ability to define envs to the vLLM containers in sample app by @nerdalert in #301

New Contributors

• @chewong made their first contribution in #327

Full Changelog: llm-d-1.0.19...llm-d-1.0.20

llm-d is a Kubernetes-native high-performance distributed LLM inference framework

What's Changed

• Sync values.schema.json with upstream k8s spec by @nerdalert in #323

Full Changelog: llm-d-1.0.18...llm-d-1.0.19

llm-d is a Kubernetes-native high-performance distributed LLM inference framework

What's Changed

• chore: update script to include logging functions / output on macOS by @cdrage in #271
• Set prometheus maximumStartupDurationSeconds to 300 by @anfredette in #281
• Added interactive prompt for enabling OpenShift user workload monitoring by @bbenshab in #280
• Slim path fix in readme by @nerdalert in #288
• add MODE variable to Istio installation script by @delavet in #284
• Script to integrate into the e2e-aws workflow by @nerdalert in #277
• Update e2e-aws to use the e2e-validation script by @vishnoianil in #279
• move & rename pd-nixl slim values file by @nerdalert in #276
• fix the installer when not using openshift, and OCP_DISABLE_INGRESS_ARGS is empty by @christian-posta in #289
• add grafana setup guide and metrics overview by @sallyom in #274
• minikube readme updates by @nerdalert in #291
• Adds Multi-Arch and Multi-OS Support to install-deps.sh Script by @danehans in #293
• Adds option "infra-only" to llmd-installer.sh #209 by @maugustosilva in #294
• feat: Adds option "download-only" to llmd-installer.sh #209 #294 by @maugustosilva in #296
• fix: install-deps.sh - handle kubectl installation on macOS without root group (#303) by @yossiovadia in #304
• For quick validation use 1st decode pod if there are multiple pods by @mnmehta in #305
• [quickstart] feat: adds a new CLI option (-g/--context) by @maugustosilva in #302
• Swap from kgateway to istio as the default in sample app by @nerdalert in #307

New Contributors

• @anfredette made their first contribution in #281
• @bbenshab made their first contribution in #280
• @delavet made their first contribution in #284
• @christian-posta made their first contribution in #289
• @danehans made their first contribution in #293
• @maugustosilva made their first contribution in #294
• @yossiovadia made their first contribution in #304
• @mnmehta made their first contribution in #305

Full Changelog: llm-d-1.0.17...llm-d-1.0.18

llm-d is a Kubernetes-native high-performance distributed LLM inference framework

What's Changed

• chore: fixes very small typo. by @cdrage in #262
• Discover the model id for the user in test-request by @nerdalert in #266
• remove dev branch refs by @Gregory-Pereira in #269

Full Changelog: llm-d-1.0.16...llm-d-1.0.17

llm-d is a Kubernetes-native high-performance distributed LLM inference framework

What's Changed

• chore: fix http helm repo publishing by @tumido in #263

Full Changelog: llm-d-1.0.15...llm-d-1.0.16

llm-d is a Kubernetes-native high-performance distributed LLM inference framework

What's Changed

• Remove pull secrets from installer by @nerdalert in #259
• chore: fix ci deps to adjust for removed dev branch by @tumido in #260
• chore: trigger release after move to public by @tumido in #261

Full Changelog: llm-d-1.0.14...llm-d-1.0.15

llm-d is a Kubernetes-native high-performance distributed LLM inference framework

What's Changed

• Fix post minikube merge prometheus install for minikube by @nerdalert in #228
• Add --minikube flag to --uninstall by @nerdalert in #224
• Add "slim" value files for each example that will run the stack on 1xL4 g6.2xlarge by @nerdalert in #230
• Add command to minikube troubleshooting section by @nerdalert in #237
• Add example of llama4 fp8 model by @cfchase in #217
• README updates for values files and minor fixes by @cfchase in #244
• Set instructions to minikube since that's tested, and vanilla kube by @cfchase in #245
• Add doc for creating a cluster for llm-d by @cfchase in #247
• Add more machine type clarity. by @nerdalert in #246
• Add blurb saying only single installation is supported by @cfchase in #249
• chore(ci): change to using public images and no extra creds in GHA by @tumido in #254
• chore(ci): fix the badge img url for release decorator by @tumido in #255
• chore: claim artifacthub ownership by @tumido in #258

Full Changelog: llm-d-1.0.13...llm-d-1.0.14