Embeddings Search

Read a single JSON record, vectorize the value at the configured query attribute using the selected provider's embedding model, and write the resulting vector back onto the same record. The output is meant to be piped straight into a downstream vector-search node (dbv::QdrantSearch, dbv::PineconeSearch, …).

The embedding model must match the model that produced the vectors in the target collection — otherwise the search returns garbage.

Pre-requisite: Install an AI provider application from Profile > {Organization} > Applications.

Parameters

Provider—REQUIRED

Configured AI application.

Model

Embedding model identifier. Must match the model used to populate the target vector collection. Defaults to the provider's recommended embedding model when left empty.

Query Attribute—REQUIRED

Field on the incoming JSON record holding the search query text. The step fails if the field is missing or non-string.

Vector Attribute—REQUIRED

Field on the outgoing JSON record where the resulting query embedding vector (array of floats) is written.

Input

JSON—REQUIRED

JSON object containing the query attribute.

Output

JSON

The same JSON object with the embedded query vector added under the configured vector attribute.