Azure Open AI RAG with Azure Search data_source

This demo shows how can we include the data source in the chat completion API itself.

How to run

• Replace the values in appsettings.json file with your environment.
  • This require deploying model in your Azure Open AI.
  • Get the Blogger API credential json file from Google cloud dashboard and save as "api-project-gkey.json" in the /src folder
• Compile and run. It will show options
• No RAG option
  • Try the option 2 to talk with Chat GPT without RAG. It can answer based on the training time around Sep 2021
• RAG option
  • First download the blog posts.
  • It will be using hardcoded tag=nas to download posts. Change that as per your blog
  • Downloaded posts will be available at /bin folder and the file name will be as per config[LocalBlogPostsFileName]
  • Next upload the documents to Azure Search Index by using Option 4. (It will be creating 'blog-index' if not present and upload to that)
  • Finally RAG using the option 5.

In case anything is missing, please create an issue or PR.

Architecture

Specifications

• .Net version - .Net 8
• Nugets referenced
  • DotNet.Helpers - To reduce some code.
  • easyconsolestd - To show menu and get input
  • Microsoft.Extensions.Hosting
  • Google.Apis.Blogger.v3 - for working with Blogger APIs
  • Azure.Search.Documents - to upload blog posts into Azure search
  • Azure.AI.OpenAI - To interact with Azure OpenAI

Dependency injection

• Supported. Refer the Program.cs file for more details
• The options are injected as dependency to the [MenuService](/src/MenuService.cs then those are invoked based on selection.

Disclaimer

This is done for learning purpose and not production ready. This require basic understanding of C# .Net 8, Azure and Open AI integration.

In case anything is missing, wrong, please raise an issue or give a PR.