Skip to content

Releases

v9.0.1

Choose a tag to compare

@github-actions github-actions released this 12 Nov 12:26

Install

Prebuilt binaries (linux amd64/arm64, macOS Apple Silicon) are
attached below, e.g.:

curl -fsSL https://github.com/weka/kvlab/releases/download/v9.0.1/kvlab_v9.0.1_$(uname -s | tr A-Z a-z)_$(uname -m | sed 's/x86_64/amd64/;s/aarch64/arm64/').tar.gz | tar -xz

Benchmark

Replay benchmark against your endpoint (see chart README for all values).
This image embeds the replay artifact, so no volume or fetch step is needed:

helm upgrade --install my-replay oci://quay.io/weka.io/helm/kvlab \
  --version v9.0.1 \
  --set endpoint=http://YOUR-LLM-HOST:8000
docker pull quay.io/weka.io/kvlab:v9.0.1

Router

Model-aware LLM router with prefix-cache affinity. Separate,
replay-less image — same binary, without the multi-GB replay layer a
router never reads.

Point it at your fleet by pod label. Each discovered pod contributes
its own port, so several DaemonSets on different ports stay one pool:

helm upgrade --install my-router oci://quay.io/weka.io/helm/kvlab-router \
  --version v9.0.1 \
  --set-string 'router.backends[0].pods.app=vllm' \
  --set-string 'router.backends[0].port=http'

A static list works the same way, and the two can be mixed:

  --set-string 'router.backends[0]=http://vllm-a:8000' \
  --set-string 'router.backends[1]=http://vllm-b:8000'

Per-model routes, a hosted fallback and credentials are in
docs/router.md.

docker pull quay.io/weka.io/kvlab-router:v9.0.1

CLI (binary is named kvlab):

go install github.com/weka/kvlab@v9.0.1

Full Changelog: v9.0.0...v9.0.1

v8.3.0

Choose a tag to compare

@github-actions github-actions released this 04 Nov 16:32

Install

Prebuilt binaries (linux amd64/arm64, macOS Apple Silicon) are
attached below, e.g.:

curl -fsSL https://github.com/weka/wekai/releases/download/v8.3.0/wekai_v8.3.0_$(uname -s | tr A-Z a-z)_$(uname -m | sed 's/x86_64/amd64/;s/aarch64/arm64/').tar.gz | tar -xz

Benchmark

Replay benchmark against your endpoint (see chart README for all values).
This image embeds the replay artifact, so no volume or fetch step is needed:

helm upgrade --install my-replay oci://quay.io/weka.io/helm/wekai \
  --version v8.3.0 \
  --set endpoint=http://YOUR-LLM-HOST:8000
docker pull quay.io/weka.io/wekai:v8.3.0

Router

Model-aware LLM router with prefix-cache affinity. Separate,
replay-less image — same binary, without the multi-GB replay layer a
router never reads.

Point it at your fleet by pod label. Each discovered pod contributes
its own port, so several DaemonSets on different ports stay one pool:

helm upgrade --install my-router oci://quay.io/weka.io/helm/wekai-router \
  --version v8.3.0 \
  --set-string 'router.backends[0].pods.app=vllm' \
  --set-string 'router.backends[0].port=http'

A static list works the same way, and the two can be mixed:

  --set-string 'router.backends[0]=http://vllm-a:8000' \
  --set-string 'router.backends[1]=http://vllm-b:8000'

Per-model routes, a hosted fallback and credentials are in
docs/router.md.

docker pull quay.io/weka.io/wekai-router:v8.3.0

CLI (binary is named wekai):

go install github.com/weka/wekai@v8.3.0

Full Changelog: v8.2.0...v8.3.0

v8.1.0

Choose a tag to compare

@github-actions github-actions released this 21 Oct 19:07

Install

Prebuilt binaries (linux amd64/arm64, macOS Apple Silicon) are
attached below, e.g.:

curl -fsSL https://github.com/weka/wekai/releases/download/v8.1.0/wekai_v8.1.0_$(uname -s | tr A-Z a-z)_$(uname -m | sed 's/x86_64/amd64/;s/aarch64/arm64/').tar.gz | tar -xz

Benchmark

Replay benchmark against your endpoint (see chart README for all values).
This image embeds the replay artifact, so no volume or fetch step is needed:

helm upgrade --install my-replay oci://quay.io/weka.io/helm/wekai \
  --version v8.1.0 \
  --set endpoint=http://YOUR-LLM-HOST:8000
docker pull quay.io/weka.io/wekai:v8.1.0

Router

Model-aware LLM router with prefix-cache affinity. Separate,
replay-less image — same binary, without the multi-GB replay layer a
router never reads.

Point it at your fleet by pod label. Each discovered pod contributes
its own port, so several DaemonSets on different ports stay one pool:

helm upgrade --install my-router oci://quay.io/weka.io/helm/wekai-router \
  --version v8.1.0 \
  --set-string 'router.backends[0].pods.app=vllm' \
  --set-string 'router.backends[0].port=http'

A static list works the same way, and the two can be mixed:

  --set-string 'router.backends[0]=http://vllm-a:8000' \
  --set-string 'router.backends[1]=http://vllm-b:8000'

Per-model routes, a hosted fallback and credentials are in
docs/router.md.

docker pull quay.io/weka.io/wekai-router:v8.1.0

CLI (binary is named wekai):

go install github.com/weka/wekai@v8.1.0

Full Changelog: v7.5.0...v8.1.0

v8.0.0

Choose a tag to compare

@github-actions github-actions released this 20 Oct 11:54

Install

Prebuilt binaries (linux amd64/arm64, macOS Apple Silicon) are
attached below, e.g.:

curl -fsSL https://github.com/weka/wekai/releases/download/v8.0.0/wekai_v8.0.0_$(uname -s | tr A-Z a-z)_$(uname -m | sed 's/x86_64/amd64/;s/aarch64/arm64/').tar.gz | tar -xz

Benchmark

Replay benchmark against your endpoint (see chart README for all values).
This image embeds the replay artifact, so no volume or fetch step is needed:

helm upgrade --install my-replay oci://quay.io/weka.io/helm/wekai \
  --version v8.0.0 \
  --set endpoint=http://YOUR-LLM-HOST:8000
docker pull quay.io/weka.io/wekai:v8.0.0

Router

Model-aware LLM router with prefix-cache affinity. Separate,
replay-less image — same binary, without the multi-GB replay layer a
router never reads.

Point it at your fleet by pod label. Each discovered pod contributes
its own port, so several DaemonSets on different ports stay one pool:

helm upgrade --install my-router oci://quay.io/weka.io/helm/wekai-router \
  --version v8.0.0 \
  --set-string 'router.backends[0].pods.app=vllm' \
  --set-string 'router.backends[0].port=http'

A static list works the same way, and the two can be mixed:

  --set-string 'router.backends[0]=http://vllm-a:8000' \
  --set-string 'router.backends[1]=http://vllm-b:8000'

Per-model routes, a hosted fallback and credentials are in
docs/router.md.

docker pull quay.io/weka.io/wekai-router:v8.0.0

CLI (binary is named wekai):

go install github.com/weka/wekai@v8.0.0

Full Changelog: v7.6.0...v8.0.0

v7.6.0

Choose a tag to compare

@github-actions github-actions released this 19 Oct 12:18

Install

Prebuilt binaries (linux amd64/arm64, macOS Apple Silicon) are
attached below, e.g.:

curl -fsSL https://github.com/weka/wekai/releases/download/v7.6.0/wekai_v7.6.0_$(uname -s | tr A-Z a-z)_$(uname -m | sed 's/x86_64/amd64/;s/aarch64/arm64/').tar.gz | tar -xz

Benchmark

Replay benchmark against your endpoint (see chart README for all values).
This image embeds the replay artifact, so no volume or fetch step is needed:

helm upgrade --install my-replay oci://quay.io/weka.io/helm/wekai \
  --version v7.6.0 \
  --set endpoint=http://YOUR-LLM-HOST:8000
docker pull quay.io/weka.io/wekai:v7.6.0

Router

Model-aware LLM router with prefix-cache affinity. Separate,
replay-less image — same binary, without the multi-GB replay layer a
router never reads.

Point it at your fleet by pod label. Each discovered pod contributes
its own port, so several DaemonSets on different ports stay one pool:

helm upgrade --install my-router oci://quay.io/weka.io/helm/wekai-router \
  --version v7.6.0 \
  --set-string 'router.backends[0].pods.app=vllm' \
  --set-string 'router.backends[0].port=http'

A static list works the same way, and the two can be mixed:

  --set-string 'router.backends[0]=http://vllm-a:8000' \
  --set-string 'router.backends[1]=http://vllm-b:8000'

Per-model routes, a hosted fallback and credentials are in
docs/router.md.

docker pull quay.io/weka.io/wekai-router:v7.6.0

CLI (binary is named wekai):

go install github.com/weka/wekai@v7.6.0

Full Changelog: v7.5.0...v7.6.0

v7.5.0

Choose a tag to compare

@github-actions github-actions released this 15 Oct 15:04

Install

Prebuilt binaries (linux amd64/arm64, macOS Apple Silicon) are
attached below, e.g.:

curl -fsSL https://github.com/weka/wekai/releases/download/v7.5.0/wekai_v7.5.0_$(uname -s | tr A-Z a-z)_$(uname -m | sed 's/x86_64/amd64/;s/aarch64/arm64/').tar.gz | tar -xz

Benchmark

Replay benchmark against your endpoint (see chart README for all values).
This image embeds the replay artifact, so no volume or fetch step is needed:

helm upgrade --install my-replay oci://quay.io/weka.io/helm/wekai \
  --version v7.5.0 \
  --set endpoint=http://YOUR-LLM-HOST:8000
docker pull quay.io/weka.io/wekai:v7.5.0

Router

Model-aware LLM router with prefix-cache affinity. Separate,
replay-less image — same binary, without the multi-GB replay layer a
router never reads.

Point it at your fleet by pod label. Each discovered pod contributes
its own port, so several DaemonSets on different ports stay one pool:

helm upgrade --install my-router oci://quay.io/weka.io/helm/wekai-router \
  --version v7.5.0 \
  --set-string 'router.backends[0].pods.app=vllm' \
  --set-string 'router.backends[0].port=http'

A static list works the same way, and the two can be mixed:

  --set-string 'router.backends[0]=http://vllm-a:8000' \
  --set-string 'router.backends[1]=http://vllm-b:8000'

Per-model routes, a hosted fallback and credentials are in
docs/router.md.

docker pull quay.io/weka.io/wekai-router:v7.5.0

CLI (binary is named wekai):

go install github.com/weka/wekai@v7.5.0

Full Changelog: v7.4.3...v7.5.0

v7.4.3

Choose a tag to compare

@github-actions github-actions released this 12 Oct 14:04

Install

Prebuilt binaries (linux amd64/arm64, macOS Apple Silicon) are
attached below, e.g.:

curl -fsSL https://github.com/weka/wekai/releases/download/v7.4.3/wekai_v7.4.3_$(uname -s | tr A-Z a-z)_$(uname -m | sed 's/x86_64/amd64/;s/aarch64/arm64/').tar.gz | tar -xz

Benchmark

Replay benchmark against your endpoint (see chart README for all values).
This image embeds the replay artifact, so no volume or fetch step is needed:

helm upgrade --install my-replay oci://quay.io/weka.io/helm/wekai \
  --version v7.4.3 \
  --set endpoint=http://YOUR-LLM-HOST:8000
docker pull quay.io/weka.io/wekai:v7.4.3

Router

Model-aware LLM router with prefix-cache affinity. Separate,
replay-less image — same binary, without the multi-GB replay layer a
router never reads.

Point it at your fleet by pod label. Each discovered pod contributes
its own port, so several DaemonSets on different ports stay one pool:

helm upgrade --install my-router oci://quay.io/weka.io/helm/wekai-router \
  --version v7.4.3 \
  --set-string 'router.backends[0].pods.app=vllm' \
  --set-string 'router.backends[0].port=http'

A static list works the same way, and the two can be mixed:

  --set-string 'router.backends[0]=http://vllm-a:8000' \
  --set-string 'router.backends[1]=http://vllm-b:8000'

Per-model routes, a hosted fallback and credentials are in
docs/router.md.

docker pull quay.io/weka.io/wekai-router:v7.4.3

CLI (binary is named wekai):

go install github.com/weka/wekai@v7.4.3

What's Changed

  • fix(router): release the half-open circuit token on a panic from attempt by @oferki in #8

New Contributors

  • @oferki made their first contribution in #8

Full Changelog: v7.4.2...v7.4.3

v7.4.2

Choose a tag to compare

@github-actions github-actions released this 05 Oct 12:04

Install

Prebuilt binaries (linux amd64/arm64, macOS Apple Silicon) are
attached below, e.g.:

curl -fsSL https://github.com/weka/wekai/releases/download/v7.4.2/wekai_v7.4.2_$(uname -s | tr A-Z a-z)_$(uname -m | sed 's/x86_64/amd64/;s/aarch64/arm64/').tar.gz | tar -xz

Benchmark

Replay benchmark against your endpoint (see chart README for all values).
This image embeds the replay artifact, so no volume or fetch step is needed:

helm upgrade --install my-replay oci://quay.io/weka.io/helm/wekai \
  --version v7.4.2 \
  --set endpoint=http://YOUR-LLM-HOST:8000
docker pull quay.io/weka.io/wekai:v7.4.2

Router

Model-aware LLM router with prefix-cache affinity. Separate,
replay-less image — same binary, without the multi-GB replay layer a
router never reads.

Point it at your fleet by pod label. Each discovered pod contributes
its own port, so several DaemonSets on different ports stay one pool:

helm upgrade --install my-router oci://quay.io/weka.io/helm/wekai-router \
  --version v7.4.2 \
  --set-string 'router.backends[0].pods.app=vllm' \
  --set-string 'router.backends[0].port=http'

A static list works the same way, and the two can be mixed:

  --set-string 'router.backends[0]=http://vllm-a:8000' \
  --set-string 'router.backends[1]=http://vllm-b:8000'

Per-model routes, a hosted fallback and credentials are in
docs/router.md.

docker pull quay.io/weka.io/wekai-router:v7.4.2

CLI (binary is named wekai):

go install github.com/weka/wekai@v7.4.2

Full Changelog: v7.4.1...v7.4.2

v7.4.1

Choose a tag to compare

@github-actions github-actions released this 28 Sep 07:54

Install

Prebuilt binaries (linux amd64/arm64, macOS Apple Silicon) are
attached below, e.g.:

curl -fsSL https://github.com/weka/wekai/releases/download/v7.4.1/wekai_v7.4.1_$(uname -s | tr A-Z a-z)_$(uname -m | sed 's/x86_64/amd64/;s/aarch64/arm64/').tar.gz | tar -xz

Benchmark

Replay benchmark against your endpoint (see chart README for all values).
This image embeds the replay artifact, so no volume or fetch step is needed:

helm upgrade --install my-replay oci://quay.io/weka.io/helm/wekai \
  --version v7.4.1 \
  --set endpoint=http://YOUR-LLM-HOST:8000
docker pull quay.io/weka.io/wekai:v7.4.1

Router

Model-aware LLM router with prefix-cache affinity. Separate,
replay-less image — same binary, without the multi-GB replay layer a
router never reads.

Point it at your fleet by pod label. Each discovered pod contributes
its own port, so several DaemonSets on different ports stay one pool:

helm upgrade --install my-router oci://quay.io/weka.io/helm/wekai-router \
  --version v7.4.1 \
  --set-string 'router.backends[0].pods.app=vllm' \
  --set-string 'router.backends[0].port=http'

A static list works the same way, and the two can be mixed:

  --set-string 'router.backends[0]=http://vllm-a:8000' \
  --set-string 'router.backends[1]=http://vllm-b:8000'

Per-model routes, a hosted fallback and credentials are in
docs/router.md.

docker pull quay.io/weka.io/wekai-router:v7.4.1

CLI (binary is named wekai):

go install github.com/weka/wekai@v7.4.1

Full Changelog: v7.4.0...v7.4.1

v7.4.0

Choose a tag to compare

@github-actions github-actions released this 16 Sep 08:49

Install

Prebuilt binaries (linux amd64/arm64, macOS Apple Silicon) are
attached below, e.g.:

curl -fsSL https://github.com/weka/wekai/releases/download/v7.4.0/wekai_v7.4.0_$(uname -s | tr A-Z a-z)_$(uname -m | sed 's/x86_64/amd64/;s/aarch64/arm64/').tar.gz | tar -xz

Benchmark

Replay benchmark against your endpoint (see chart README for all values).
This image embeds the replay artifact, so no volume or fetch step is needed:

helm upgrade --install my-replay oci://quay.io/weka.io/helm/wekai \
  --version v7.4.0 \
  --set endpoint=http://YOUR-LLM-HOST:8000
docker pull quay.io/weka.io/wekai:v7.4.0

Router

Model-aware LLM router with prefix-cache affinity. Separate,
replay-less image — same binary, without the multi-GB replay layer a
router never reads.

Point it at your fleet by pod label. Each discovered pod contributes
its own port, so several DaemonSets on different ports stay one pool:

helm upgrade --install my-router oci://quay.io/weka.io/helm/wekai-router \
  --version v7.4.0 \
  --set-string 'router.backends[0].pods.app=vllm' \
  --set-string 'router.backends[0].port=http'

A static list works the same way, and the two can be mixed:

  --set-string 'router.backends[0]=http://vllm-a:8000' \
  --set-string 'router.backends[1]=http://vllm-b:8000'

Per-model routes, a hosted fallback and credentials are in
docs/router.md.

docker pull quay.io/weka.io/wekai-router:v7.4.0

CLI (binary is named wekai):

go install github.com/weka/wekai@v7.4.0

Full Changelog: v7.3.0...v7.4.0