AI Charts

AI Charts is a collection of Helm Charts for installing and managing mainstream AI applications on Kubernetes.

Currently Supported Applications

Dify

Some code was referenced from https://github.com/douban/charts/blob/master/charts/dify

Dify is an LLM application development platform that supports the entire process from creating prototypes to deployment.

Installing Dify

1. Add the Helm repository:

helm repo add ai-charts https://magicsong.github.io/ai-charts/
helm repo update

2. Create a custom values.yaml file:

global:
    host: "mydify.example.com"
    enableTLS: false

    image:
        # Set to the latest version of Dify
        # Check versions here: https://github.com/langgenius/dify/releases
        tag: "1.0.1"
    
    extraBackendEnvs:
    - name: SECRET_KEY
      value: "please-generate-your-own-key"
    - name: LOG_LEVEL
      value: "INFO"
    - name: VECTOR_STORE
      value: "milvus"

ingress:
    enabled: true
    className: "nginx"

3. Install the chart:

helm install dify ai-charts/dify -f values.yaml

4. Required - Run database migrations after installation or the instance won't work:

# Run migrations
kubectl exec -it <dify-pod-name> -- flask db upgrade

Upgrading Dify

Modify global.image.tag in your values.yaml to the desired version, then run:

helm upgrade dify ai-charts/dify -f values.yaml

Required - Run database migrations after the upgrade:

kubectl exec -it <dify-pod-name> -- flask db upgrade

Laminar

Laminar is a comprehensive AI platform with multiple components including PostgreSQL, RabbitMQ, ClickHouse, Qdrant vector database, semantic search services, and more.

Installing Laminar

1. Add the Helm repository:

helm repo add ai-charts https://magicsong.github.io/ai-charts/
helm repo update

2. Create a custom values.yaml file:

# Basic configuration
frontend:
  ingress:
    enabled: true
    className: nginx
    hosts:
      - host: laminar.example.com
        paths:
          - path: /
            pathType: Prefix
  env:
    sharedSecretToken: "your-secure-token"
    nextauthUrl: "https://laminar.example.com"
    nextauthSecret: "your-nextauth-secret"
    nextPublicUrl: "https://laminar.example.com"

# Data persistence configuration
postgresql:
  auth:
    username: postgres
    password: your-secure-password
    database: laminar
  persistence:
    enabled: true
    size: 8Gi

rabbitmq:
  auth:
    username: user
    password: your-secure-password
  persistence:
    enabled: true
    size: 8Gi

clickhouse:
  auth:
    username: default
    password: your-secure-password
  persistence:
    enabled: true
    size: 10Gi

qdrant:
  persistence:
    enabled: true
    size: 8Gi

# Semantic search configuration
semanticSearchService:
  env:
    cohereApiKey: "your-cohere-api-key"

3. Install the chart:

helm install laminar ai-charts/laminar -f values.yaml

Customizing Laminar

Laminar supports various customization options:

• External PostgreSQL: Configure postgresql.external.enabled=true and related connection parameters
• Resource Configuration: Each component can be configured with standard Kubernetes resource settings
• Persistence: All data components (PostgreSQL, RabbitMQ, ClickHouse, Qdrant) support persistent storage
• Ingress: The frontend component can be exposed via Ingress

For more configuration options, please refer to the values.yaml file in the Helm Chart.

Contributing

Pull requests and issues are welcome to help improve this project.

License

This project is licensed under the Apache License 2.0.