Introduction

The T-Systems Smartchat is an AI chatbot application. Smartchat enables secure interaction with your own data sources using natural language. You can search documents for information, summarize their content, and generate new content or specific output formats based on them.

The technical foundation is a Retrieval Augmented Generation (RAG) architecture. RAG retrieves information from files or databases and provides it to the language model (Large Language Model, LLM) as context. This means chat responses are grounded in the retrieved content. Smartchat displays the name of each source used.

Data privacy

The language models used in Smartchat are not trained on user inputs. Data processing takes place in a secure cloud environment within the EU, operated by Telekom and hosted in T-Cloud Public. A selection of LLMs is available and continuously updated with the latest models.

Smartchat is built on a fork of OpenWebUI combined with a custom multi-tenant Enterprise RAG framework, with content organized in isolated tenants called Spaces.

What Smartchat adds to OpenWebUI

Multi-Tenancy

Fully isolated tenants (Spaces) with separate users, documents, and configurations — all on a single deployment.

Enterprise RBAC

Fine-grained role-based access control across spaces, admin panels, and API endpoints — synced with your SSO provider.

Async Ingestion Pipeline

Scalable, fully asynchronous document ingestion — handles large document volumes without blocking.

Security Hardening

Sandboxed code execution, input validation, and attack surface reduction beyond upstream defaults.

Compliance & Accessibility

Certified for enterprise compliance standards and accessibility requirements.

Confluence Integration

Connect Confluence Data Center spaces — pages are auto-synced and become searchable context in chats.

Disclaimer

Please note that this documentation covers only the generic features of Smartchat. Since the feature set provided is always individually defined in your contract, not all features described here may be part of your deployment.

Additionally, custom features or role permissions may have been agreed upon that go beyond the standard scope of this documentation.