The`Pipeline` above generates embeddings and tsvectors for the`abstract` and splits and generates embeddings and tsvectors for the`text`.

We can now perform search over both the`text` and`abstract` key of our documents. See the[guide for vector search](vector-search) for more information on how to do this.

**This section is only relevant for self hosted instances of PostgresML**. These parameters are never required for instances hosted by PostgresML.

Some HuggingFace models require the argument`trust_remote_code=true`. To enable this, pass it as a parameter in the pipeline construction:

pipeline= Pipeline(

"v0",

{

"text": {

"semantic_search": {

"model":"Alibaba-NLP/gte-base-en-v1.5",

"parameters": {

"trust_remote_code":True

}

}

}

}

)

Pass your HuggingFace token into the pipeline to access gated repos:

pipeline= Pipeline(

"v0",

{

"text": {

"semantic_search": {

"model":"Alibaba-NLP/gte-base-en-v1.5",

"parameters": {

"trust_remote_code":True,

"token":"YOUR_TOKEN"

}

}

}

}

)