Azure AI Search - Upsert
Category: Generative AIPackage: TruBot.OpenAI.Activities
Purpose
To upsert an Azure AI Search vector database.
Description
The Azure AI Search - Upsert component upserts an Azure AI Search vector database with the vector embedding representing real-world data like document content. It first send the data to an embedding model that generates the embedding related to the data. It then calls an endpoint at which the vector database service is available to upsert the embedding. The embedding are stored in an index whose name is passed to the endpoint along with the embedding. The embedding stored in the index is in a form that allows to find the best matching response for a user query.
In the given example, the component activity upserts the vector database by first sending the data represented by the IndexContent to the embedding model represented by AzureOpenAICOnnection. The vector embedding that is received in response is sent to the vector database by calling the required endpoint. Along with the embedding, the name of the index that is to be created i.e., AzureSearchIndex, is also sent to the endpoint.
Properties
|
Property |
Data Type |
Description |
|
General |
||
|
Display Name |
String |
Display name of the component. |
|
Enable Pause |
-NA- |
Option to pause the activity (related to the component) during a job execution after receiving a pause control signal from the Cockpit application. The checkbox is selected by default. |
|
Input |
|
|
|
API Key |
String |
The API key is used to authenticate access to the embedding model. (See Embedding connection property.) |
|
IEmbedding |
An object that is used to secure access to an AI/ML service running on one of the AI/ML platforms. It also provides the name of the embedding model that is to be used for performing the AI/ML activities. Note: The object can be generated by calling one of the activities available under the Generative AI > Embedding category. |
|
|
Endpoint |
String |
The URL at which the vector store service for inserting the embedding is available. It is called by the activity after the embedding is received from the embedding model. |
|
Index Content |
Contents |
An object containing the content that need to be upserted in the vector database. Note: Use the Create Vector Content activity to create the required object. |
|
Index Name |
String |
The name of the index in which the embedding is to be stored in a form that enables to find the best matching response for a user query. |
|
Splitter |
ITextsplitter |
Name of the splitter, in case a splitter is required. Splitters are used to chunk data into a smaller size so that the token-size limit of the embedding model is met. |
|
Misc |
|
|
|
Enable Bookmark |
-NA- |
Option to set a bookmark. |
|
Is Reserved |
-NA- |
Option to disable data tracing related to the component. |
|
Note: The property names marked with the * sign are the mandatory properties. |
||
