Azure OpenAI RAG + memory Demo app

Infrastructure

Obviously, that uses Azure infrastructure. You will need to create following resources (please, follow the sequence):

1. Azure OpenAI. This is our core, which will be replying to all the questions we have.
   1. Create chat completion deployment, that will be used for conversations. GPT-4o-mini was used in this case, but feel free to experiment with others.
   2. Create embedding deployment, that will be used for generating embeddings in RAG. text-embedding-ada-002 was used in this case, but feel free to experiment with others.
2. Azure Storage Account, and we will store RAG files here:
   1. You also need to create Container there.
3. Azure AI Search, this will be our search enginer for RAG:
   1. No worries, you can start with Free tier.
4. Document Intelligence, this will analyze files and extract text from them:
   1. Free tier is good enough for testing purposes.

Running locally

If you want to test it, you need two things:

1. Install docker
2. Create file, for example .demo.env and add the following content there:

AZURE_OPENAI_CHATGPT_DEPLOYMENT=<name of your chat completion deployment from 1.1>
AZURE_OPENAI_EMBEDDING_DEPLOYMENT=<name of your embedding deployment from 1.2>
AZURE_OPENAI_ENDPOINT=<endpoint of your Azure Open AI deployment from 1>
AZURE_OPENAI_API_KEY=<key for using your Azure Open AI deployment from 1>
AZURE_STORAGE_CONNECTION_STRING=<conenction string for Azure Storage from 2>
AZURE_STORAGE_CONTAINER_NAME=<container where to store files from 2.1>
AZURE_SEARCH_SERVICE_ENDPOINT=<endpoint of Azure AI Search from 3>
AZURE_SEARCH_SERVICE_API_KEY=<key for Azure AI Search from 3>
AZURE_SEARCH_INDEX_NAME=<name of the index you want to be automatically created>
AZURE_FORMRECOGNIZER_SERVICE_ENDPOINT=<endpoint for Document Intelligence from 4>
AZURE_FORMRECOGNIZER_SERVICE_API_KEY=<key for Document Intelligence from 4>

Then you can do:

docker compose -f docker-compose.yml --env-file .demo.env build
docker compose -f docker-compose.yml --env-file .demo.env up -d

Application becomes available at http://localhost. Enjoy!

Developing locally

Backend

As code is written in C#, use Visual Studio (Community Edition is more than enough). Prior running it, you will need to setup the same environment variables as in running locally step.

Then run it as http. It should open with Swagger on port 5247.

Frontend

As this part is heavy on TypeScript (React, Redux, Ant Design), I recommend using Visual Studio Code. Depending on port used for Backend instance, you might want to adjust BACKEND_HOST_URL in ./build-configuration/dev.properties file. First, you need to do:

yarn

Then start it with:

yarn start

Application becomes available at http://localhost:8080.