Hi everyone,
I’m running into an issue where my collection is being created with text2vec-palm
as the vectorizer, even though I’ve set everything to use text2vec-google (Gemini embeddings).
vectorizer_config = wvc.config.Configure.Vectorizer.text2vec_google(
project_id=“myid”,
)
Docker Compose setup
volumes:
- weaviate_gemini:/var/lib/weaviate
- ./credentials/google_credentials.json:/etc/weaviate/google_credentials.json:ro
restart: on-failure:0
environment:
QUERY_DEFAULTS_LIMIT: 5000
AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED: "true"
PERSISTENCE_DATA_PATH: "/var/lib/weaviate"
ENABLE_API_BASED_MODULES: true
DEFAULT_VECTORIZER_MODULE: text2vec-google
ENABLE_MODULES: text2vec-google
CLUSTER_HOSTNAME: "node1"
MODULES_CLIENT_TIMEOUT: "1h"
USE_GOOGLE_AUTH: "true"
GOOGLE_APPLICATION_CREDENTIALS: "/etc/weaviate/google_credentials.json"
Actual schema being created
{
"class": "Collection_ddd",
"invertedIndexConfig": {
"bm25": { "b": 0.75, "k1": 1.2 },
"cleanupIntervalSeconds": 60,
"stopwords": { "preset": "en" }
},
"moduleConfig": {
"text2vec-palm": {
"modelId": "gemini-embedding-001",
"projectId": "***removed***",
"vectorizeClassName": true
}
},
"properties": [
{
"name": "fileId",
"dataType": ["text"],
"moduleConfig": {
"text2vec-palm": { "skip": true, "vectorizePropertyName": true }
}
},
{
"name": "fileContent",
"dataType": ["text"],
"moduleConfig": {
"text2vec-palm": { "skip": false, "vectorizePropertyName": true }
}
}
],
"vectorIndexConfig": { "distance": "cosine" },
"vectorIndexType": "hnsw",
"vectorizer": "text2vec-palm"
}
Even though I’m using text2vec-google
in both my Docker config and collection creation code, the schema always ends up with text2vec-palm
as the vectorizer.
When I try to query, I get this error:
no vectorizer module “text2vec-palm” is configured
Versions
-
Weaviate server:
1.27.4
-
Weaviate Python client:
4.10.1