Configuration

KM service requires a configuration file to start, typically named appsettings.Development.json for local development or appsettings.Production.json for production environments. The appropriate configuration file is automatically loaded, based on the ASPNETCORE_ENVIRONMENT environment variable. If your development workstation doesn’t have this environment variable, it’s recommended to create it: ASPNETCORE_ENVIRONMENT == “Development”. In a production environment you’ll want to set ASPNETCORE_ENVIRONMENT = “Production”.

The KM repository includes a setup wizard to help you create your initial appsettings.Development.json file:

cd service/Service
dotnet run setup

Follow the on-screen prompts to create or edit the configuration file. If needed, you can interrupt the script and run it multiple times to modify the settings.

Run the web service (upload and search endpoints)? YES

Protect the web service with API Keys? NO —— you can leave this off for this test

Enable OpenAPI swagger doc at /swagger/index.html? YES

Run the .NET pipeline handlers as a service? YES

How should memory ingestion be orchestrated? Using asynchronous distributed queues

Which queue service will be used? SimpleQueues —— this will use volatile queues in memory, suitable only for tests

  Directory where to store queue messages: _tmp_queues

Where should the service store files? SimpleFileStorage —— this will use volatile storage in memory, suitable only for tests. You can manually edit appsettings.development.json to persist files on disk.

  Directory where to store files: _tmp_files

Which service should be used to extract text from images? None

When importing data, generate embeddings? YES

When searching for text and/or answers, which embedding generator should be used? Azure OpenAI or OpenAI

  OpenAI <text/chat model name> [current: gpt-3.5-turbo-16k]:      Press ENTER to use default
  OpenAI <embedding model name> [current: text-embedding-ada-002]: Press ENTER to use default
  OpenAI <API Key>: sk-*********************************

When searching for answers, which memory DB service contains the records? SimpleVectorDb —— this will use volatile storage in memory, suitable only for tests. You can manually edit appsettings.development.json to persist files on disk, or choose one of the available options suggested.

  Directory where to store vectors: _tmp_vectors

When generating answers and synthetic data, which LLM text generator should be used? Azure OpenAI or OpenAI

Log level? Information

Great! If you completed the wizard, you should be ready to start the service and run the examples below.

• If you selected any of the “simpleXYZ” dependencies, then data will be stored in memory only, and automatically discarded when the service stops. Edit the configuration file manually to persist data on disk. More information here.
• The configuration wizard uses some default settings that you might want to change. After running the examples, take a look at the included appsettings.json to see all the available options, and read the Service Configuration doc for more information.

Next: Start the service