RAG Platform¶
The RAG Platform is the retrieval-augmented generation (RAG) backend developed as part of the KI:Expertisezentrum.nrw. It aims to provide client applications with an interface for RAG functionality.
Client applications upload documents into collections. They then ask questions that are answered from those documents, with citations that point back to the sources.
The platform is application-agnostic. It speaks an OpenAI-style HTTP API and works with any OpenAI-compatible inference provider.
Docs version
These pages describe RAG Platform 1.1.x, the product release rag-docker 1.1.0. Use the version selector in the header to switch between releases.
Choose your path¶
Deployment admins install the platform, configure it, and make it available to client applications in their network.
- Install with Docker Compose or with Podman quadlets
- Create tenants and API keys
- Configuration reference
- Upgrade to 1.1
Client developers build applications against the public /v1 API of a running
platform instance. You need its base URL and an API key from the operator.
- Run a local instance for development
- API quickstart
- Streaming and citations
- API reference
Regardless of the role, you will benefit from reading the concepts.