Hey there,
In the original flow in the sample application in the repo, /api/search was wired to the Ottoman query builder example, which generates SQL++ and not vector or full-text queries. Without a queryType, .build() just returns an empty string, that’s expected for that path.
I updated the repository example to be more straightforward going forward though. The updated flow now routes /api/search through the performSearch service. This is the part of the repo that uses the Couchbase SDK’s scope.search() API with a VectorSearch request, which is the intended way to run vector searches in this sample.
For it to return meaningful results, you’ll still need:
- Documents in your Capella bucket/scope with an
embeddingarray field - A Capella Vector Search index on that field
This sample is meant to illustrate how vector search fits into an app’s flow, not to be a complete production setup. If you want to take it further, the Couchbase docs walk through the full Capella + Node.js SDK setup:
With the current controller in place, once your index and embeddings are ready, /api/search will execute a proper vector search against Capella and return ranked results.
All the best,
Ben