How to use text-embedding-3-small in Docker

Support
1

Hi community,

Currently, I tried to use text-embedding-3-small by configuring docker-compose file. But by default, text2vec-openai using text-embedding-ada-002.
Is there a way that I can set to use text-embbeding-3-small in docker-compose ? Here is my docker-compose.yml:

version: '3.4'
services:
  weaviate:
    command:
    - --host
    - 0.0.0.0
    - --port
    - '8080'
    - --scheme
    - http
    image: semitechnologies/weaviate:1.23.10 # Adjust version if needed
    ports:
    - 8080:8080
    - 50051:50051
    volumes:
    - weaviate_data:/var/lib/weaviate
    restart: on-failure:0
    environment:
      ## general
      # ORIGIN: https://my-weaviate-deployment.com
      QUERY_DEFAULTS_LIMIT: 25
      QUERY_MAXIMUM_RESULTS: 10000     
      PERSISTENCE_DATA_PATH: '/var/lib/weaviate'
      AUTOSCHEMA_ENABLED: true
      ENABLE_MODULES: 'text2vec-transformers,text2vec-openai'
      # DEFAULT_VECTORIZER_MODULE: 'text2vec-transformers'      
      DEFAULT_VECTORIZER_MODULE: 'text2vec-openai'      
      ## Module-specific
      TRANSFORMERS_INFERENCE_API: http://t2v-transformers:8080
      ## Authentication
      AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED: 'true'
      # Enables API key authentication.
      # AUTHENTICATION_APIKEY_ENABLED: 'true'
      # List one or more keys, separated by commas. Each key corresponds to a specific user identity below.
      # AUTHENTICATION_APIKEY_ALLOWED_KEYS: 'this-is-a-secret-key-for-me, this-is-a-secret-key-for-admin'
      # List one or more user identities, separated by commas. Each identity corresponds to a specific key above.
      # AUTHENTICATION_APIKEY_USERS: 'this-is-me, this-is-admin'
      ## Authorization
      # AUTHORIZATION_ADMINLIST_ENABLED: 'true'
      # AUTHORIZATION_ADMINLIST_USERS: 'this-is-admin'
      # AUTHORIZATION_ADMINLIST_READONLY_USERS: 'this-is-me'
      ## Multi-node setups
      CLUSTER_HOSTNAME: 'node1'
      # CLUSTER_JOIN: weaviate-node-1:
    env_file:
      - .env

  t2v-transformers:
    image: semitechnologies/transformers-inference:sentence-transformers-multi-qa-MiniLM-L6-cos-v1
    environment:
      ENABLE_CUDA: 0 # set to 1 to enable
      # NVIDIA_VISIBLE_DEVICES: all # enable if running with CUDA
volumes:
  weaviate_data:

Thank you for your help!

2

hi @cong.dao !!

You need to explicitly set in at Collection create time.

Like so:

collection = client.collections.create(
    "Collection",
    vectorizer_config=wvc.config.Configure.Vectorizer.text2vec_openai(
        model="text-embedding-3-small"
    ),
)

AFAIK, there isn’t an environment variable to change a module/integration default from docker compose.

Please, if you think this could be a good feature, feel free to open a feature request in Issues · weaviate/weaviate · GitHub

Thanks!

1 Like