Versions (relevant - OpenSearch/Dashboard/Server OS/Browser): 2.13
Describe the issue: We are trying to create an ingest pipeline and index for our hybrid search cluster, but the documents we are trying to ingest have a mapping of nested fields that we need to create embeddings for. How am I suppose to setup my ingest pipeline and index such that I can create embeddings for these nested fields?
Here is an example schema of what our items look like:
{
“id” 123,
“_source”: {
“attributes” :{“name”: “test”}
}
}
And here is the pipeline configuration we are trying:
PUT /_ingest/pipeline/item-new
{
“description”: “A text embedding pipeline”,
“processors”: [
{
“text_embedding”: {
“model_id”: “model”,
“field_map”: {
“attributes.name”: “name_embedding”
}
}
}
]
}
Please let us know how we can create this properly such that we can get these embeddings to be generated for nested fields.