Configure the Deployment

With the cluster ready, use values.yaml to configure Cosine. This file becomes the source of truth for what’s running in your environment.

Image tags

Set the tag for each service to the Cosine version you want to run so every component stays in lockstep:

image:
  repository: registry.cosine.enterprises/azure/api
  pullPolicy: IfNotPresent
  tag: REPLACE_WITH_IMAGE_TAG

Repeat for dashboard, api, ccode, and parser.

LLM configuration

Provide the inference endpoint and the model you want to use. These settings control the default model Cosine will reach for:

- name: LITELLM_BASE_URL
  value: "http://RESOURCE_NAME.openai.azure.com/openai/v1"
- name: LITELLM_GENIE_MODEL_OVERRIDE
  value: "gpt-5.2"
- name: LITELLM_API_KEY
  valueFrom:
    secretKeyRef:
      name: llm-auth
      key: token

Domain configuration

Set the dashboard and API URLs, then update the ingress hostnames so the UI and backend agree on their public address:

dashboard:
  env:
    - name: DASHBOARD_URL
      value: "https://REPLACE_WITH_YOUR_DOMAIN"
    - name: API_URL
      value: "https://REPLACE_WITH_YOUR_DOMAIN/api"
    - name: NEXT_PUBLIC_API_URL
      value: "https://REPLACE_WITH_YOUR_DOMAIN/api"
ingress:
  dashboardHostname: DOMAIN
  tls:
    - secretName: domain-tls
    hosts:
      - DOMAIN
env:
  - name: API_URL
    value: "https://REPLACE_WITH_YOUR_DOMAIN/api"
  - name: DASHBOARD_URL
    value: "https://REPLACE_WITH_YOUR_DOMAIN"
  - name: COORS_ORIGIN
    value: "https://REPLACE_WITH_YOUR_DOMAIN"
  - name: DOMAIN
    value: "REPLACE_WITH_YOUR_DOMAIN"

Login secrets

Set JWT_SECRET and COOKIE_SECRET to 32‑character hex strings. You can generate them with:

openssl rand -hex 32

Then update values.yaml:

- name: JWT_SECRET
  value: "REPLACE_WITH_VALUE"
- name: COOKIE_SECRET
  value: "REPLACE_WITH_VALUE"