llama-server helm

No introduction found. Create it?

Install

Install with:

helm repo add app-template oci://ghcr.io/bjw-s-labs/charts/
helm install llama-server app-template/app-template -f values.yaml

Examples

See examples from other people.

Top Repositories (2 out of 3)

NameRepoStarsVersionTimestamp
llama-serverjoryirving/home-ops2165.0.17 hours ago
llama-serverjoryirving/home-ops2165.0.17 hours ago

Values

See the most popular values for this chart:

KeyTypes
service.app.ports.http.port (3)
8080
number
service.app.ports.http.targetPort (1)
8080
number
service.app.annotations."lbipam.cilium.io/ips" (2)
10.69.10.20
string
service.app.externalTrafficPolicy (2)
Cluster
string
service.app.type (2)
LoadBalancer
string
service.app.controller (1)
app
string
controllers.llama-server.annotations."reloader.stakater.com/auto" (2)
true
string
controllers.llama-server.containers.app.args[] (2)
- --host
- "0.0.0.0"
- --port
- "8080"
- --alias
- self-hosted
- --model
- /models/gemma4-26b/gemma-4-26B-A4B-it-UD-Q5_K_XL.gguf
- --mmproj
- /models/gemma4-26b/mmproj-F16.gguf
- --ctx-size
- "262144"
- --n-gpu-layers
- "99"
- --flash-attn
- on
- --parallel
- "3"
- --cont-batching
- -sps
- "0.90"
- --no-context-shift
- --cache-prompt
- --kv-unified
- --ctx-checkpoints
- "4"
- --checkpoint-every-n-tokens
- "16384"
- --image-min-tokens
- "1120"
- --image-max-tokens
- "1120"
- --metrics
- --temp
- "0.4"
- --top-p
- "0.92"
- --top-k
- "40"
- --min-p
- "0"
- --batch-size
- "4096"
- --ubatch-size
- "2048"
- --threads
- "12"
- --threads-batch
- "16"
- --no-mmap
string
controllers.llama-server.containers.app.command[] (2)
- llama-server
string
controllers.llama-server.containers.app.image.pullPolicy (2)
Always
string
controllers.llama-server.containers.app.image.repository (2)
docker.io/kyuz0/amd-strix-halo-toolboxes
string
controllers.llama-server.containers.app.image.tag (2)
rocm-7.2.3@sha256:a07e7576a800a8f77cd6ae1a4522d6c7e005f49cf047470a5183a2e1a4031135
string
controllers.llama-server.containers.app.probes.liveness.custom (2)
true
boolean
controllers.llama-server.containers.app.probes.liveness.enabled (2)
true
boolean
controllers.llama-server.containers.app.probes.liveness.spec.failureThreshold (2)
6
number
controllers.llama-server.containers.app.probes.liveness.spec.httpGet.path (2)
/health
string
controllers.llama-server.containers.app.probes.liveness.spec.httpGet.port (2)
8080
number
controllers.llama-server.containers.app.probes.liveness.spec.initialDelaySeconds (2)
30
number
controllers.llama-server.containers.app.probes.liveness.spec.periodSeconds (2)
30
number
controllers.llama-server.containers.app.probes.readiness.custom (2)
true
boolean
controllers.llama-server.containers.app.probes.readiness.enabled (2)
true
boolean
controllers.llama-server.containers.app.probes.readiness.spec.failureThreshold (2)
6
number
controllers.llama-server.containers.app.probes.readiness.spec.httpGet.path (2)
/health
string
controllers.llama-server.containers.app.probes.readiness.spec.httpGet.port (2)
8080
number
controllers.llama-server.containers.app.probes.readiness.spec.initialDelaySeconds (2)
20
number
controllers.llama-server.containers.app.probes.readiness.spec.periodSeconds (2)
10
number
controllers.llama-server.containers.app.probes.startup.custom (2)
true
boolean
controllers.llama-server.containers.app.probes.startup.enabled (2)
true
boolean
controllers.llama-server.containers.app.probes.startup.spec.failureThreshold (2)
360
number
controllers.llama-server.containers.app.probes.startup.spec.httpGet.path (2)
/health
string
controllers.llama-server.containers.app.probes.startup.spec.httpGet.port (2)
8080
number
controllers.llama-server.containers.app.probes.startup.spec.periodSeconds (2)
10
number
controllers.llama-server.containers.app.resources.limits.memory (2)
80Gi
string
controllers.llama-server.containers.app.resources.requests.cpu (2)
1
number, string
controllers.llama-server.containers.app.resources.requests.memory (2)
20Gi
string
controllers.llama-server.containers.app.securityContext.privileged (2)
true
boolean
controllers.llama-server.initContainers.model-download.command[] (2)
- /bin/sh
- -ec
- set -euo pipefail if [ ! -s /models/gemma4-26b/gemma-4-26B-A4B-it-UD-Q5_K_XL.gguf ] || [ ! -s /models/gemma4-26b/mmproj-F16.gguf ]; then pip install --no-cache-dir "huggingface_hub[hf_transfer]" mkdir -p /models/gemma4-26b rm -rf /models/gemma4-26b/.cache/huggingface/download/gemma-4-26B-A4B-UD-Q5_K_XL.gguf.lock hf download \ unsloth/gemma-4-26B-A4B-it-GGUF \ gemma-4-26B-A4B-it-UD-Q5_K_XL.gguf \ mmproj-F16.gguf \ --local-dir /models/gemma4-26b echo "Gemma-4-26B-A4B-Q5_K_XL download complete" else echo "Gemma-4-26B-A4B-Q5_K_XL already downloaded, skipping" fi
string
controllers.llama-server.initContainers.model-download.env.HF_HUB_ENABLE_HF_TRANSFER (2)
1
string
controllers.llama-server.initContainers.model-download.envFrom[].secretRef.name (2)
huggingface
string
controllers.llama-server.initContainers.model-download.image.repository (2)
python
string
controllers.llama-server.initContainers.model-download.image.tag (2)
3.14-slim
string
controllers.llama-server.initContainers.model-download.resources.limits.memory (2)
10Gi
string
controllers.llama-server.initContainers.model-download.resources.requests.cpu (2)
250m
string
controllers.llama-server.initContainers.model-download.resources.requests.memory (2)
512Mi
string
controllers.app.annotations."reloader.stakater.com/auto" (1)
true
string
controllers.app.containers.app.args[] (1)
- --host
- "0.0.0.0"
- --port
- "8080"
- --models-preset
- /app/models.ini
- --models-max
- "1"
- --parallel
- "1"
- --metrics
- --slots
string
controllers.app.containers.app.env.LLAMA_CACHE (1)
/cache
string
controllers.app.containers.app.env.TZ (1)
${TIMEZONE}
string
controllers.app.containers.app.image.repository (1)
ghcr.io/ggml-org/llama.cpp
string
controllers.app.containers.app.image.tag (1)
server-cuda13@sha256:1b3694b9a8f68b95e4667c8b598c3e46b69dc2702ab227ebe2b2bd2282bb34d6
string
controllers.app.containers.app.probes.liveness.custom (1)
true
boolean
controllers.app.containers.app.probes.liveness.enabled (1)
true
boolean
controllers.app.containers.app.probes.liveness.spec.failureThreshold (1)
3
number
controllers.app.containers.app.probes.liveness.spec.httpGet.path (1)
/health
string
controllers.app.containers.app.probes.liveness.spec.httpGet.port (1)
8080
number
controllers.app.containers.app.probes.liveness.spec.initialDelaySeconds (1)
30
number
controllers.app.containers.app.probes.liveness.spec.periodSeconds (1)
30
number
controllers.app.containers.app.probes.liveness.spec.timeoutSeconds (1)
5
number
controllers.app.containers.app.probes.readiness.custom (1)
true
boolean
controllers.app.containers.app.probes.readiness.enabled (1)
true
boolean
controllers.app.containers.app.probes.readiness.spec.failureThreshold (1)
3
number
controllers.app.containers.app.probes.readiness.spec.httpGet.path (1)
/health
string
controllers.app.containers.app.probes.readiness.spec.httpGet.port (1)
8080
number
controllers.app.containers.app.probes.readiness.spec.initialDelaySeconds (1)
5
number
controllers.app.containers.app.probes.readiness.spec.periodSeconds (1)
10
number
controllers.app.containers.app.probes.readiness.spec.timeoutSeconds (1)
3
number
controllers.app.containers.app.probes.startup.custom (1)
true
boolean
controllers.app.containers.app.probes.startup.enabled (1)
true
boolean
controllers.app.containers.app.probes.startup.spec.failureThreshold (1)
12
number
controllers.app.containers.app.probes.startup.spec.httpGet.path (1)
/health
string
controllers.app.containers.app.probes.startup.spec.httpGet.port (1)
8080
number
controllers.app.containers.app.probes.startup.spec.initialDelaySeconds (1)
5
number
controllers.app.containers.app.probes.startup.spec.periodSeconds (1)
5
number
controllers.app.containers.app.probes.startup.spec.timeoutSeconds (1)
3
number
controllers.app.containers.app.resources.limits.cpu (1)
8
number
controllers.app.containers.app.resources.limits.memory (1)
26Gi
string
controllers.app.containers.app.resources.limits."nvidia.com/gpu" (1)
1
number
controllers.app.containers.app.resources.requests.cpu (1)
1
number
controllers.app.containers.app.resources.requests.memory (1)
6Gi
string
controllers.app.containers.app.resources.requests."nvidia.com/gpu" (1)
1
number
defaultPodOptions.hostIPC (2)
true
boolean
defaultPodOptions.nodeSelector."topology.kubernetes.io/gpus" (2)
amd
string
defaultPodOptions.tolerations[].effect (2)
NoSchedule
string
defaultPodOptions.tolerations[].key (2)
llm-workload
string
defaultPodOptions.tolerations[].operator (2)
Exists
string
defaultPodOptions.affinity.nodeAffinity.requiredDuringSchedulingIgnoredDuringExecution.nodeSelectorTerms[].matchExpressions[].key (1)
nvidia.com/gpu
string
defaultPodOptions.affinity.nodeAffinity.requiredDuringSchedulingIgnoredDuringExecution.nodeSelectorTerms[].matchExpressions[].operator (1)
In
string
defaultPodOptions.affinity.nodeAffinity.requiredDuringSchedulingIgnoredDuringExecution.nodeSelectorTerms[].matchExpressions[].values[] (1)
- true
string
defaultPodOptions.runtimeClassName (1)
nvidia
string
persistence.dev-dri.globalMounts[].path (2)
/dev/dri
string
persistence.dev-dri.hostPath (2)
/dev/dri
string
persistence.dev-dri.type (2)
hostPath
string
persistence.dev-kfd.globalMounts[].path (2)
/dev/kfd
string
persistence.dev-kfd.hostPath (2)
/dev/kfd
string
persistence.dev-kfd.type (2)
hostPath
string
persistence.models.existingClaim (2)
llm-models
string
persistence.models.globalMounts[].path (2)
/models
string
persistence.cache.enabled (1)
true
boolean
persistence.cache.existingClaim (1)
llama-server
string
persistence.cache.globalMounts[].path (1)
/cache
string
persistence.config.globalMounts[].path (1)
/app/models.ini
string
persistence.config.globalMounts[].readOnly (1)
true
boolean
persistence.config.globalMounts[].subPath (1)
models.ini
string
persistence.config.name (1)
llama-server-config
string
persistence.config.type (1)
configMap
string
route.app.hostnames[] (1)
- {{ .Release.Name }}.${SECRET_DOMAIN}
string
route.app.parentRefs[].name (1)
envoy-internal
string
route.app.parentRefs[].namespace (1)
network
string
route.app.rules[].backendRefs[].identifier (1)
app
string
route.app.rules[].backendRefs[].port (1)
80
number