AI Search semantic config for RAG option

[dabedin]

Using the RAG option, code relies on AISEARCH_SEMANTIC_CONFIG env variable. Is there guidance on how to configure the semantic configuration to add on which index?

[msalemor]

The easiest way to get this configuration is to:

• Deploy an Azure storage account and deploy the source documents to the blob container
• Deploy Azure AI Search
• Deploy OpenAI text-embedding model
  • Navigate to Azure AI Search and click on "Integrated Vectorization"
  • this step will ask for the container source, and text embedding location
  • it will also ask if you want to enable Semantic Ranker, do this
  • The previous step creates a number of skills and starts the process of ingestion
  • At the end, the URL should be: https://[NAME].search.windows.net/indexes/[INDEX_NAME]/docs/search?api-version=2024-05-01-preview
  • The name of the semantic config will be: [INDEX_NAME]-semantic-configuration

[dabedin]

Yet not clear which source document are expected to be uploaded in the flow of the demo.