Unable to connect Azure AI search to Knowledge of an Agent

Hello Alekhya, sorry for the late response. I have followed your advice with Ingredients - EDM.String

Title - EDM.String

URL - EDM.String

and its still crashing everytime I try to proceed with the connector.

Hi Patrick Martin Castro
Even though the schema uses Edm.string, the actual data can have unexpected values- such as arrays, objects, or even zero values- which can cause indexer to fail.

sometimes, special characters (especially from scraped or user-related materials) can break the parsing unexpectedly.

can cause timeout or crash during very long wire or large record field sequencing.

Hi Patrick Martin Castro, We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet. In case if you have any resolution, please do share that same with the community as it can be helpful to others. Otherwise, will respond with more details and we will try to help.

Hi @Alekhya Vaddepally , still having the same issue, no progress.

Hello @Patrick Martin Castro , Azure AI Search is primarily used when you need to search through a large collection of documents. This could be anything from text documents to other types of content indexed in the Azure AI Search service. Currently, you must create vectorized indexes using integrated vectorization. and then use in knowledge source. Refer this document https://learn.microsoft.com/en-us/microsoft-copilot-studio/knowledge-azure-ai-search

https://learn.microsoft.com/en-us/azure/search/vector-search-how-to-configure-vectorizer

Refer to this doc for using azure-search-documents and the Azure SDK for Python, applying data chunking and vectorization in an indexer pipeline.

In the OpenAI resource, have you added the Cognitive Services OpenAI User role? Have you included the sample "gpt-4" module before adding "text-embedding-ada-002" using this DOC

I tried to create vectorized indexes and it would succesfully connect but i would have errors when making chats to the vectorized index when connected to Azure openai due to quota errors. So im trying to find other ways to connect my openai to the copilot. Another problem I encountered is that I couldn't find my connection here even though I have established a connection in Power Apps.

My roles for this resource is Azure Ai Developer, Cognitive Services OpenAI Contributor

and Contributor

Hello @Patrick Martin Castro ,

Verify the quota limits for your Azure OpenAI resource. You can do this in the Azure portal under the "Usage + quotas" section of your OpenAI resource. Reduce the number of queries or the complexity of requests to stay within quota limits. For example, use batching or caching to minimize API calls.

refer this links for Use SharePoint content for generative answers

https://learn.microsoft.com/en-us/microsoft-copilot-studio/nlu-generative-answers-sharepoint-onedrive

https://learn.microsoft.com/en-us/microsoft-copilot-studio/nlu-generative-answers-azure-openai

I was able to connect the search service to the Copilot now. All i needed to do was to vectorize the index. The index alone being imported was causing the error.

Hello @Patrick Martin Castro ,

Refer to this doc for using azure-search-documents and the Azure SDK for Python, applying data chunking and vectorization in an indexer pipeline.

the Storage Blob Data Reader role to storage
In the OpenAI resource, have you added the Cognitive Services OpenAI User role? Have you included the sample "gpt-4" module before adding "text-embedding-ada-002" using this DOC

We need to have the OpenAI resources and supported models.

Check the supported provider and models here.

Since you already having the models and embedding endpoints with in ai services you can use either sdk or rest api,

below is the sample to create indexer on json array data via python sdk

from azure.core.credentials import AzureKeyCredential
from azure.search.documents.indexes import SearchIndexClient, SearchIndexerClient
from azure.search.documents.indexes.models import (  
SearchIndexer,  
IndexingParameters,  
IndexingParametersConfiguration,  
SearchIndexerDataSourceConnection,  
SearchIndexerDataContainer,  
SearchIndexerSkillset,  
FieldMapping,  
InputFieldMappingEntry,  
OutputFieldMappingEntry,  
AzureOpenAIEmbeddingSkill,  
)
 
# Azure Search Service Configuration
service_name = "ai_search_name"
index_name = "index_name"
api_key = "your-azure-search-api-key"
endpoint = f"https://{service_name}.search.windows.net/"
 
# Azure Blob Storage Configuration
blob_connection_string = "your-blob-connection-string"
container_name = "aicsvdata"
 
# Azure OpenAI Configuration
azure_openai_service = "your-openai-service-name"
azure_openai_api_key = "your-openai-api-key"
azure_openai_embedding_deployment = "text-embedding-ada-002"
 
def create_indexer_with_skillset(index_client, indexer_client):
    """Creates an indexer with a skillset to vectorize the 'plot' field."""
    data_source_name = "blob-datasource"
    indexer_name = "blob-indexer-with-vector"
    skillset_name = "plot-vectorization-skillset"
 
    # 1. Data Source
    container = SearchIndexerDataContainer(name=container_name)
    data_source = SearchIndexerDataSourceConnection(
        name=data_source_name,
        connection_string=blob_connection_string,
        container=container,
        type="azureblob",
    )
 
    # 2. Skillset - Embedding Generation
    embedding_skill = AzureOpenAIEmbeddingSkill(
        name="plot-embedding",
        description="Generates vector embeddings for plot field",
        context="/document",
        resource_url=f"https://{azure_openai_service}.openai.azure.com/",
        api_key=azure_openai_api_key,
        deployment_name=azure_openai_embedding_deployment,
        model_name="text-embedding-ada-002",
        inputs=[InputFieldMappingEntry(name="text", source="/document/plot")],
        outputs=[OutputFieldMappingEntry(name="embedding", target_name="PlotVector")],
    )
 
    skillset = SearchIndexerSkillset(
        name=skillset_name,
        description="Skillset to vectorize plot",
        skills=[embedding_skill],
    )
 
    # 3. Field Mappings
    field_mappings = [
        FieldMapping(source_field_name="/document/PlotVector", target_field_name="PlotVector"),
    ]
	indexing_parameters = IndexingParameters(
        query_timeout=None,
        configuration=IndexingParametersConfiguration(
            parsing_mode="jsonArray",  # Options: 'default', 'delimitedText', 'json', 'jsonArray', 'jsonLines', 'text', 'markdown'
            data_to_extract="contentAndMetadata"  # Options: 'allMetadata', 'contentAndMetadata', 'storageMetadata'
        )
    )
 
    # 4. Indexer
    indexer = SearchIndexer(
        name=indexer_name,
        data_source_name=data_source_name,
        target_index_name=index_name,
        skillset_name=skillset_name,
        parameters = indexing_parameters,
        output_field_mappings = field_mappings
    )
 
    # Create skillset and data source
    indexer_client.create_or_update_skillset(skillset)
    indexer_client.create_or_update_data_source_connection(data_source)
 
    # Create and run the indexer
    result = indexer_client.create_or_update_indexer(indexer)
    print(f"Indexer '{result.name}' created.")
    indexer_client.run_indexer(indexer_name)
 
def main():
    try:
        credential = AzureKeyCredential(api_key)
        index_client = SearchIndexClient(endpoint, credential)
        indexer_client = SearchIndexerClient(endpoint, credential)
 
        create_indexer_with_skillset(index_client, indexer_client)
 
    except Exception as ex:
        print(f"An error occurred: {ex}")
 
if __name__ == "__main__":
    main()

Hello @Patrick Martin Castro , Since the original issue about being unable to connect Azure AI search to Knowledge of an Agent is solved, can you tell me what resolved it? I also have comments on your additional question. Could I close this question by posting the answer ?

What worked was I needed to vectorize data first. Import and vectorize data instead of import data alone.