The repository contains a Blazor server application built with DevExpress XAF. It connects to Azure SQL Database, leverages native vector search, and integrates Azure OpenAI to embed documents and provide chat-based answers. Documents are uploaded, tokenized into chunks, and stored with embedding vectors; questions are reformulated and matched against these embeddings to retrieve relevant text. The project demonstrates how to combine DevExpress AI tools, EF Core’s vector search capabilities, and Semantic Kernel services to implement retrieval-augmented generation within a XAF application.
This simple PoC project XAF Blazor uses native vector support from SQL Azure Vector functions, Native Vector Support in Azure SQL Database and Vector Support in Azure SQL Database
- .NET 9
- Visual Studio 2022 (or latest preview)
- Restore packages:
dotnet restore. - Update
appsettings.jsonconnection string and Azure OpenAI keys as described. - Build and run the server project:
dotnet run --project XAFVectorSearch.Blazor.Server.
Set the database connection in appsettings
{
"ConnectionStrings": {
"DefaultConnection": "Server=.;Database=VectorSearch;Trusted_Connection=True;MultipleActiveResultSets=true"
}
}In the AzureOpenAI section of the appsettings.json file, insert the access key for Microsoft's OpenAI service.
Once the document is loaded, click on Embedding document to generate the vector of the document (embedding model).
When chatting, a sorting operation will be performed based on the cosine similarity between the vectors of the document chunks in db and the vector of the entered phrase, and the top MaxRelevantChunks (in appsettings.json) will be taken to be used in the autocompletion context.
