In continuation of the previous two posts to vectorize and analyze drone images and to perform semantic retrieval for a wide variety of drone sensing applications, this article lists some of the errors encountered in setting up the infrastructure on Azure and the possible resolutions:

1. The index size appears as 0 even though the index works correctly to list entries. The exact error message is:

400: An error occurred when calling Azure Cognitive Search: Azure Search Error: 400, message='Server responded with status 400. Error message: {"error":{"code":"","message":"This index must have valid semantic configurations defined before using the 'semanticConfiguration' query parameter.\r\nParameter name: semanticConfiguration"}}', url='https://srch-vision-01.search.windows.net/indexes/index00/docs/search?api-version=2024-03-01-preview'

Semantic search with Azure Search failed with configuration=default.

Please ensure semantic search is enabled and you have semantic search quota available on your instance.

  This error can be resolved by going to the vector search settings of the Azure AI Search resource and ensuring that the semantic search configuration is created or updated followed by resetting and restarting the indexer. Semantic configurations describe the title, content, and keywords fields that will be used for semantic ranking, captions, highlights, and answers.

2. SubscriptionIsOverQuotaForSku error when compute cannot be provisioned to deploy a chat user interface web application. The exact error message is:

This region has quota of 0 instances for your subscription. Try selecting different region or SKU.

The resolution for this error involves increasing the limits defined on my quotas page.

3. Error encountered when adding an embedding model to the vector store:

Exact error message is:

Error: Request failed with status code 400

    at ht (https://ai.azure.com/assets/manualChunk_data-fetch-84262679.js:14:85613)

    at dt (https://ai.azure.com/assets/manualChunk_data-fetch-84262679.js:14:85796)

    at XMLHttpRequest.S (https://ai.azure.com/assets/manualChunk_data-fetch-84262679.js:15:1655)

azureml://registries/azure-openai/models/text-embedding-ada-002/versions/2

Failed to updated Index "index00", error: "The request is invalid. Details: definition: Error with vectorizer 'vectorizer-1748496106073': 'dimensions' parameter is out of range for the 'modelName' value of 'text-embedding-ada-002'. Value was '1024', Expected value to be 1536."

The dimensions for the vectors inserted into the vector store is typically 1024 but the embedding models operate on vectors with dimensions as shown:

• text-embedding-ada-002 (1536 dimensions)

• text-embedding-3-small (1536 dimensions)

• text-embedding-3-large (3072 dimensions)

And increasing the dimensions of the vectors in the store from 1024 to those corresponding to your choice of embedding model resolves this error. This must be followed by resetting and rerunning the indexer for the vectors especially if the vectorizer is added to the vector profiles leveraging Hnsw or ExhaustiveKnn Vector algorithms. This can be done with the following cli commands:

• POST /indexers/[indexer name]/reset?api-version=[api-version]

• POST /indexers/[indexer name]/run?api-version=[api-version]

• POST /indexers/[indexer name]/status?api-version=[api-version]

For example:

POST {endpoint}/indexers('{indexerName}')/search.run?api-version=2024-07-01

#Codingexercise: https://1drv.ms/w/c/d609fb70e39b65c8/Echlm-Nw-wkggNYlIwEAAAABD8nSsN--hM7kfA-W_mzuWw?e=jr5Kc9