docs/integrations/vectorstores/azuresearch/

[giscus[bot]]

docs/integrations/vectorstores/azuresearch/

Azure AI Search (formerly known as Azure Search and Azure Cognitive Search) is a cloud search service that gives developers infrastructure, APIs, and tools for information retrieval of vector, keyword, and hybrid queries at scale.

https://python.langchain.com/docs/integrations/vectorstores/azuresearch/

[RyuseiTakahashi]

Thank you for taking the time to read my inquiry.

I have a question about the following library.
from langchain_community.vectorstores.azuresearch import AzureSearch

Let's say I've defined two SearchFieldDataType.Collection(SearchFieldDataType.Single) columns (vector fields) in one index.

Is it possible to perform a cosine similarity vector search on these two SearchFieldDataType.Collection(SearchFieldDataType.Single) columns (vector fields)?

Example:
Let's say I've set the following index schema.

field name	type
content	SearchFieldDataType.String
content_vector	SearchFieldDataType.Collection(SearchFieldDataType.Single)
title	SearchFieldDataType.String
title_vector	SearchFieldDataType.Collection(SearchFieldDataType.Single)

Is it possible to perform a similarity search on both vectors in the fields "content_vector" and "title_vector" and obtain the data with the highest similarity score?

Below is a sample program.

I would like to perform a similarity search on the title_vector and content_vector in the result variable below to obtain the data with the highest similarity.

How can I do this?

fields = [
    SimpleField(
        name="id",
        type=SearchFieldDataType.String,
        key=True,
        filterable=True,
    ),
    SearchableField(
        name="content",
        type=SearchFieldDataType.String,
        searchable=True,
    ),
    SearchField(
        name="content_vector",
        type=SearchFieldDataType.Collection(SearchFieldDataType.Single),
        searchable=True,
        vector_search_dimensions=len(embeddings.embed_query("test text")),
        vector_search_profile_name="vector-profile-myplayground-20241018172801",
    ),
    SearchableField(
        name="title",
        type=SearchFieldDataType.String,
        searchable=True,
    ),
    SearchField(
        name="title_vector",
        type=SearchFieldDataType.Collection(SearchFieldDataType.Single),
        searchable=True,
        vector_search_dimensions=len(embeddings.embed_query("test text")),
        vector_search_profile_name="vector-profile-myplayground-20241018172801",
    ),
]

vector_store: AzureSearch = AzureSearch(
    azure_search_endpoint=vector_store_endpoint,
    azure_search_key=vector_store_api_key,
    index_name="my-index-name",
    embedding_function=embeddings.embed_query,
    fields=fields,
)

result = vector_store.similarity_search_with_relevance_scores(
    query=question,
    k=5,
    score_threshold=0.60
)

Thank you very much for your assistance and time.
Sincerely

[davide-evangelisti-prom]

Hey LangChain,

I'm encountering an issue with the vector_store.delete(document) method in the Azure AI Search integration. Although the method returns True, the document isn't actually removed from the index. Any insights on how to fix this?

Thanks!

[rupeshsharma28]

Hi,
I am creating a vector store and retrieving the top documents for a query.
I want to extract Answer confidence Score along with Semantic Reranker Score of top documents. Answer fields are basically the top results generated by Azure AI Search along with its confidence score.
I am using below function to get reranker score:

vectorstore.semantic_hybrid_search_with_score(query,k=5,score_type='reranker_score')

But I don't see any option of extracting Answer Confidence Score in LangChain. Is there a way to extract that?