62 lines
1.5 KiB
YAML
62 lines
1.5 KiB
YAML
---
|
|
# tasks file for localai
|
|
- name: Add skynet repo
|
|
kubernetes.core.helm_repository:
|
|
name: go-skynet
|
|
repo_url: https://go-skynet.github.io/helm-charts/
|
|
register: repo_update
|
|
|
|
- name: Update Helm repos
|
|
command: helm repo update
|
|
when: repo_update.changed
|
|
|
|
- name: Deploy LocalAI
|
|
kubernetes.core.helm:
|
|
name: localai
|
|
chart_ref: go-skynet/local-ai
|
|
release_namespace: localai
|
|
create_namespace: true
|
|
values:
|
|
deployment:
|
|
image:
|
|
tag: latest-gpu-nvidia-cuda-12
|
|
runtimeClassName: nvidia
|
|
secretEnv:
|
|
- name: LOCALAI_DISABLE_WEBUI
|
|
value: "true"
|
|
- name: LOCALAI_API_KEY
|
|
value: "{{ localai_api_keys | join(',') }}"
|
|
- name: LOCALAI_WATCHDOG_IDLE
|
|
value: "true"
|
|
- name: LOCALAI_WATCHDOG_IDLE_TIMEOUT
|
|
value: "5m"
|
|
- name: LOCALAI_WATCHDOG_BUSY
|
|
value: "true"
|
|
resources:
|
|
limits:
|
|
nvidia.com/gpu: 1
|
|
persistence:
|
|
models:
|
|
size: 256Gi
|
|
output:
|
|
size: 128Gi
|
|
tolerations:
|
|
- key: specialResources
|
|
operator: Equal
|
|
value: GPU
|
|
effect: NoSchedule
|
|
ingress:
|
|
enabled: true
|
|
className: nginx
|
|
annotations:
|
|
cert-manager.io/cluster-issuer: ca-issuer
|
|
hosts:
|
|
- host: localai.eom.dev
|
|
paths:
|
|
- path: /
|
|
pathType: ImplementationSpecific
|
|
tls:
|
|
- secretName: localai-tls
|
|
hosts:
|
|
- localai.eom.dev
|