v0.0.3

What's Changed

• Refactoring deployment scripts for Kubernetes and Kind by @WheelyMcBones in #226
• Add prom queries for avg input tokens and ttft by @vishakha-ramani in #194
• Enhance Helm chart security configuration and add development values by @mamy-CS in #242
• Refactoring deployment scripts and integrating llm-d-inference-sim in emulated deployment environment by @WheelyMcBones in #237

Full Changelog: v0.0.2...v0.0.3

v0.0.2

What's Changed

• Update crd docs by @asm582 in #184
• Update README.md to include ref to new home for container image by @clubanderson in #183
• Update ci-release.yaml to do multi arch by @clubanderson in #182
• Removing unsupported architectures from image multi-arch build by @WheelyMcBones in #192
• Setting minNumReplicas to 1 by default by @WheelyMcBones in #189
• Update ci-release.yaml to include 'latest' tag by @clubanderson in #193
• Update QuickStart documentation to disable Rosetta on Apple Silicon by @myechuri in #195
• Fix zero divide in queue analyzer by @atantawi in #190
• add helm chart for ocp, supporting files, and instructions to install wva, prom adaptor, and llm-d in simple form by @clubanderson in #200
• removed 'oc adm policy' (replaced with clusterrolebinding) and rollout commands - no longer needed by @clubanderson in #207
• Deployment script and readme for wva + llmd on openshift by @mamy-CS in #203
• updated files to consume llmd model name and modelID by @clubanderson in #209
• Fixing tolerance function by @WheelyMcBones in #213
• Remove resource accounting logic and disable limited mode by @mamy-CS in #210
• E2es on openshift using sharegpt data by @mamy-CS in #201
• remove epp from wva helm chart and instructions to delete it after wv… by @clubanderson in #215
• Adding unit tests for the internal optimizer code by @WheelyMcBones in #204
• Add metrics validation and health monitoring system with Kubernetes conditions by @mamy-CS in #214
• refactor: Reorganize repository structure and documentation by @mamy-CS in #216
• Changing OCP deployment script to use the Helm chart by @WheelyMcBones in #212
• Documentation update by @mamy-CS in #223
• Parameterizing OCP script by @WheelyMcBones in #218
• Optimize VariantAutoscaling's owner setting by @learner0810 in #219
• refactor: Externalize metric names and labels to constants package by @ev-shindin in #228
• Enhancements to OC E2E Testing by @Vezio in #220
• Helm Chart Refactoring by @Vezio in #222

New Contributors

• @myechuri made their first contribution in #195
• @learner0810 made their first contribution in #219
• @Vezio made their first contribution in #220

Full Changelog: v0.0.1...v0.0.2

v0.0.1

What's Changed

• Add link to prerequisites in readme by @atantawi in #3
• Fix references in docs and readme files by @atantawi in #8
• Have sample demo data in a common repo by @atantawi in #9
• Remove control loop by @atantawi in #10
• Changes to move to new design by @asm582 in #11
• change interface of optimizer by @asm582 in #13
• add logger with additional code changes by @asm582 in #15
• Initial integration of inferno model analyzer and optimizer functionality by @atantawi in #17
• feat: install cluster with multiple nodes, gpus by @haroldship in #21
• Fix prometheus address by @haroldship in #25
• Inferno emulator mode by @mamy-CS in #18
• Revert "Inferno emulator mode" by @mamy-CS in #27
• Vendor vllme (new-metric branch) into inferno-autoscaler by @mamy-CS in #24
• automated inferno deployment for dev by @mamy-CS in #28
• add license file by @mamy-CS in #31
• Scale down variant to one replica when no traffic by @atantawi in #33
• Fix setupwithmanager in controller by @asm582 in #29
• use ctrl runtime backed cache for listing nodes by @asm582 in #26
• resolve merge issue by @asm582 in #34
• Variant to keep accelerator by @atantawi in #30
• add retries by @asm582 in #35
• enable HA in controller by @asm582 in #36
• improve error handling and retries for current design by @asm582 in #37
• update reconciler to smaller, composable helper functions and reduce inline logic by @asm582 in #38
• Align vllme metrics with vllm for autoscaler compatibility by @vishakha-ramani in #32
• rem modelservice requeue for reconcile by @asm582 in #40
• Support configured maximum batch size by @atantawi in #41
• Remove hard-coded accelerator name by @atantawi in #43
• llm-d integration by @WheelyMcBones in #42
• Actuator Emit custom metrics to Prometheus by @mamy-CS in #39
• Remove requeue when optimization fails by @asm582 in #49
• return errors for cm config by @asm582 in #53
• Deployment and test env cleanup by @mamy-CS in #50
• update readme by @mamy-CS in #54
• simplify watchandrun loop by @asm582 in #56
• Fix llmd integration by @WheelyMcBones in #58
• move to openAI modelid format by @asm582 in #57
• Use reconciler to run periodically by @asm582 in #59
• Update readme and install cm in inferno ns by @asm582 in #61
• multi-arch build by @mamy-CS in #62
• Modified vllme load generator by @vishakha-ramani in #60
• loadgen deterministic mode doc update by @mamy-CS in #63
• fixed waiting for Gateway and EPP deployments by @WheelyMcBones in #64
• Add documentation for modeling and analysis by @atantawi in #67
• Report allocatable resources by @asm582 in #66
• Update modelling documentation by @atantawi in #69
• enabling installation for amd64 arch by @WheelyMcBones in #68
• x86 llmd infra installation fixes by @mamy-CS in #72
• improve logging readability by @WheelyMcBones in #77
• refactoring backoff logic into global backoff by @WheelyMcBones in #89
• align slo names to community terms by @asm582 in #101
• first basic E2E tests by @WheelyMcBones in #90
• fix lint errors by @asm582 in #105
• E2E scaling tests by @WheelyMcBones in #104
• fix make test by @asm582 in #108
• Add optimize section by @asm582 in #109
• Remove dummy analyzer code by @asm582 in #110
• add crd api docs by @asm582 in #111
• fixes for crd and docs by @asm582 in #113
• fix shortname issue by @asm582 in #115
• add tls configuration by @mamy-CS in #103
• Testing continuous generated load and multiple VAs scenarios by @WheelyMcBones in #112
• add gha workflows by @clubanderson in #121
• remove precommit by @clubanderson in #124
• Fix llm-d deploy and modify tests to align with community feedback by @WheelyMcBones in #122
• add make test-e2e by @clubanderson in #126
• ignore sync errors from zap logger by @asm582 in #125
• Handle infeasible optimization solution by @atantawi in #102
• Fix E2E test execution on CI by @WheelyMcBones in #130
• Remove manual trigger logic by @asm582 in #128
• Add unit tests to Inferno-autoscaler components by @WheelyMcBones in #133
• Changing E2E to check emitted Inferno metrics by @WheelyMcBones in #135
• HPA integration by @WheelyMcBones in #137
• Integrating llm-d infra into E2E tests by @WheelyMcBones in #138
• Add api unit tests by @asm582 in #140
• Include HPA in README config by @WheelyMcBones in #142
• consolidate yaml samples in markdown file by @vishakha-ramani in #141
• Computing scaling decision based on ratio metric for HPA by @WheelyMcBones in #143
• Changes to the collector: token query by @vishakha-ramani in #146
• Upgrade optimizer-light to v0.5.0 by @atanta...