Lama-Index with AI Search

[Gianlucq]

Good evening here some questions relative to the use of llama-index with Azure AI Search (ex Azure Cognitive Search)

1. Using object CognitiveSearchVectorStore is possible to get also searchable fields? Or only filterable field? If yes, how?

2. Is possible also to have more than one "embedings" column? or just one?

Regards!

[dosubot[bot]]

🤖

Hi @Gianlucq! 👋 I'm Dosu, a friendly bot here to assist you with the LlamaIndex repository while we're waiting for a real human maintainer. I'm here to help you solve bugs, answer your questions, and guide you to become a contributor. How's your day going?

Sure, I can help with that.

1. Yes, the CognitiveSearchVectorStore object in Llama-Index can be used to retrieve both searchable and filterable fields. This is done through the _create_metadata_index_fields method, which creates index fields for storing metadata values and sets the filterable attribute to True. Here's a snippet of the relevant code:

field = SimpleField(name=field_name, type=index_field_type, filterable=True)
index_fields.append(field)

You can find this in the llama_index/vector_stores/cogsearch.py file.

2. As for your second question, it is not possible to have more than one "embeddings" column when using Llama-Index with Azure AI Search. This is because the _create_index method in the CognitiveSearchVectorStore class creates a single "embedding" field in the Azure Cognitive Search index. Here's the relevant code:

SearchField(
    name=self._field_mapping["embedding"],
    type=SearchFieldDataType.Collection(SearchFieldDataType.Single),
    hidden=False,
    searchable=True,
    filterable=False,
    sortable=False,
    facetable=False,
    vector_search_dimensions=self.embedding_dimensionality,
    vector_search_configuration="default",
)

This field is used to store the embedding vector for each document in the index. The field is set to be searchable but not filterable, sortable, or facetable. The dimensionality of the embedding vector is specified by the embedding_dimensionality attribute of the CognitiveSearchVectorStore instance.

I hope this answers your questions. If you have any more, feel free to ask!

Best regards,
Dosu

Sources

• llama_index/vector_stores/cogsearch.py

---

This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

If you want to continue the conversation, start your reply with @dosu-bot.

[Gianlucq]

I was trying to implement your example on how to create AI Search Index and store vector store index data embedding on it.
I abled to do this:
image

But when make some queries seems it does not work:

image

How can solve it?

[Gianlucq]

I was trying to implement your example on how to create AI Search Index and store vector store index data embedding on it.
I abled to do this:

But when make some queries seems it does not work:

How can solve it?