Documentation

Routing for Search and Query Scenarios

Our Retrieval Suite needs to have the following mandatory routes to function well at search time.

Database

Routes

  1. Suite → Database Server

Explanation

For productive scenarios, our Suite must have access to a SQL database, i.e., either Microsoft SQL, Postgres SQL or Oracle. The database is used to store configurations and the internal state.

Authentication

Routes

  1. Suite → Entra Id

  2. Browser → Entra Id

Explanation

The administration interface is a registered Entra Id application and users must be able to execute OIDC flows against Entra Id, when accessing the administration interface. This does not mean that the administration interface is exposed to the internet. But the browser, e.g. on the same VM, must be allowed to execute this flow. Also the Suite must be able to access https://graph.microsoft.com to verify the user identity and token.

Search Interface and Query Pipelines

Routes

  1. Clients → Entra Id

  2. Clients → Suite

  3. Suite → Entra Id (https://graph.microsoft.com)

  4. Suite → Search Engine

  5. (Optional) Suite → LLM

Explanation

If you use our search interface, users must be able to connect to our RheinInsights Retrieval Suite. The authentication is based on Entra Id and thus the clients must be able to perform an OIDC flow. Also the Suite itself must be allowed to reach out to Entra Id through the Graph APIs.

If you enable our retrieval augmented generation flows, the Suite must be able to access the LLM, i.e., either your Azure Open AI deployment or a local LLM.

And of course, the Suite must be allowed to query the search engine.

Update Checks

Not mandatory is the following route.

Route

  1. Suite → https://www.rheininsights.com

Explanation

Our Suite performs update checks every once in a while Update Check . The API endpoint is part of our public website.