Hi everyone,
I’m working through the “Semantic Cache with LangChain” technical challenge on the Capella Free Tier. My cluster is confirmed to be running Couchbase Server 7.6.x .
I have successfully used a Python script to add vector embeddings (384-dim) to the hotel
documents in the travel-sample
bucket.
My goal is to create a vector index to use the VECTOR_SEARCH()
function in a SQL++ query. However, when I try to create the index using the documented syntax, it fails.
DDL Command:
CREATE INDEX ix_hotel_embedding_vector
ON `travel-sample`.inventory.hotel(description_embedding)
WITH {"index_type": "vector", "dimensions": 384, "similarity": "cosine"};
The Error: Whether I run this command in the Capella Query Editor, or send it directly to the Query Service API (port 18093) using curl
or Postman, I consistently receive the same error:
{
"code": 5000,
"msg": "GSI CreateIndex() - cause: Invalid parameters in with-clause: 'index_type'. Valid parameters are 'nodes', 'defer_build', ...",
"status": "fatal"
}
This error suggests that the Query Service on this build does not recognize the index_type
parameter, which is required for creating a GSI vector index.
Additional Context: My initial approach was to use the Search Service (FTS). However, I was blocked because the Capella Free Tier UI does not provide the necessary fts_searcher
or Bucket Admin
roles required for my application user to query the Search index.
My Question: Given that the server version is 7.6.x, is the creation of GSI vector indexes disabled or restricted on the Capella Free Tier? Or is there an alternative