Come caricare un documento vettorizzato su ai search usando REST API?

Andrea Rizzo 0 Reputation points
2024-05-24T15:22:36.61+00:00

Ciao a tutti, sto cercando di creare un Rag ed ho un dubbio: ho un documento e l'ho suddiviso in chunks, embeddato i singoli chunks con ada openai e adesso ho quindi i vettori dei singoli chunks. Creato l'indice etc.., devo caricare i documenti e sto usando come body:

{

    "value": [

        {

            "@search.action": "upload",

            "file_id": "d*********3",

            "file_name": "D*****************************e.pdf",

            "file_description": "*******************",

            "group_ids": ["********************"],

            "Tags": ["2023"],

            "DescriptionVector": ....... ,

            "ContentVector": ............

        }

]

}

Il description vector non mi dà problemi, il dubbio è sul content vector in quanto non sto capendo come passare tutti i chunks.-

Edited: Translated into English

How to upload a vectorized document to ai search using REST API?

Hi everyone, I'm trying to create a Rag and I have a doubt: I have a document and I divided it into chunks, embedded the individual chunks with ada openai and now I have the vectors of the individual chunks. Once the index etc. has been created, I need to load the documents and I'm using as body:

{

"value": [

    {

        "@search.action": "upload",

        "file_id": "d*********3",

        "file_name": "D*****************************e.pdf",

        "file_description": "********************",

        "group_ids": ["********************"],

        "Tags": ["2023"],

        "DescriptionVector": ....... ,

        "ContentVector": ............

    }
]

}



The description vector doesn't give me any problems, the doubt is about the content vector as I'm not understanding how to pass all the chunks.
Azure AI Search
Azure AI Search
An Azure search service with built-in artificial intelligence capabilities that enrich information to help identify and explore relevant content at scale.
1,065 questions
{count} votes

Your answer

Answers can be marked as Accepted Answers by the question author, which helps users to know the answer solved the author's problem.