Deploying a different database

This pattern supports several types of vector databases, EDB Postgres for Kubernetes, Elasticsearch, Redis, Microsoft SQL Server, and the cloud-deployed Azure SQL Server. By default the pattern will deploy EDB Postgres for Kubernetes as a vector database. To use a different vector database, change the global.db.type parameter to ELASTIC, MSSQL etc. in your local branch in values-global.yaml.

global:
  pattern: rag-llm-gitops
  options:
    useCSV: false
    syncPolicy: Automatic
    installPlanApproval: Automatic
  # Possible values for RAG vector DB db.type:
  #   REDIS    -> Redis (Local chart deploy)
  #   EDB      -> PGVector (Local chart deploy)
  #   ELASTIC  -> Elasticsearch (Local chart deploy)
  #   MSSQL    -> MS SQL Server (Local chart deploy)
  #   AZURESQL -> Azure SQL (Pre-existing in Azure)
  db:
    index: docs
    type: EDB
  # Models used by the inference service (should be a HuggingFace model ID)
  model:
    vllm: ibm-granite/granite-3.3-8b-instruct
    embedding: sentence-transformers/all-mpnet-base-v2

  storageClass: gp3-csi

main:
  clusterGroupName: hub
  multiSourceConfig:
    enabled: true
    clusterGroupChartVersion: 0.9.*

This is also where you are able to update both the LLM model served by the vLLM inference service as well as the embedding model used by the vector database.

Edit this page Open a documentation issue