Step 1: Configure CodeMie API and LiteLLM Values

This step is required for both Automated and Manual setups.

Configure CodeMie API Values

Add the following environment variables to your CodeMie API values.yaml: codemie-api/values-<cloud>.yaml

extraEnv:
  - name: LLM_PROXY_MODE
    value: 'lite_llm'
  - name: LLM_PROXY_ENABLED
    value: 'true'
  - name: LITE_LLM_URL
    value: 'http://litellm.litellm:4000'
  - name: LITE_LLM_TAGS_HEADER_VALUE
    value: 'global'
  - name: LITE_LLM_APP_KEY
    valueFrom:
      secretKeyRef:
        name: litellm-integration
        key: litellm-app-key
  - name: LITE_LLM_MASTER_KEY
    valueFrom:
      secretKeyRef:
        name: litellm-integration
        key: litellm-master-key

Configure LiteLLM Values

Configure litellm/values-<cloud_name>.yaml

litellm-helm:
  ingress:
    enabled: true
    annotations: {}
    className: 'nginx'
    hosts:
      - host: litellm.%%DOMAIN%% # Replace with your domain
        paths:
          - path: /
            pathType: ImplementationSpecific
    tls: []

  redis:
    global:
      # Storage Class Configuration
      defaultStorageClass: 'your-storage-class' # e.g., "gp3", "standard", etc.
    enabled: false # Redis is disabled by default

Enable Database Initialization Job

dbInitJob automatically creates the postgres_litellm database and user during Helm deployment. It is disabled by default.

dbInitJob:
  enabled: true
  initImage: alpine/psql:18.3
  pgAdminSecret:
    name: codemie-postgresql  # Secret with PostgreSQL admin credentials
    userKey: PG_USER
    passwordKey: PG_PASS

info

When enabled, the deployment script (helm-charts.sh) automatically creates the codemie-postgresql secret using admin credentials from deployment_outputs.env — either CODEMIE_POSTGRES_DATABASE_USER/CODEMIE_POSTGRES_DATABASE_PASSWORD (shared database) or LITELLM_POSTGRES_DATABASE_USER/LITELLM_POSTGRES_DATABASE_PASSWORD (dedicated LiteLLM database). For manual deployment, create this secret before running Helm — see Manual Deployment.

Redis Configuration

By default, Redis deployment is disabled (enabled: false). Redis is an optional component that enhances LiteLLM proxy capabilities for specific use cases.

When to Enable Redis

Set enabled: true if you need any of the following features:

1. Response Caching

Cache LLM API responses to reduce costs and latency. Redis stores responses for identical queries, enabling faster response times and lower API costs.

Learn more: LiteLLM Caching Documentation

2. Distributed Rate Limiting and Load Balancing

When running multiple LiteLLM proxy instances (e.g., multiple Kubernetes pods), Redis synchronizes rate limits and load balancing state across all instances. This ensures consistent TPM/RPM enforcement and proper request distribution.

Learn more: LiteLLM Load Balancing Documentation

Next Steps

Continue to Cloud Provider Authentication.