Internal AI assistant for a social services provider

In the daily work of a social-services organisation, employees often need to determine quickly which internal procedure applies to a specific situation—for example, after an accident involving a company vehicle or when handling a particular client situation.

The required information exists in internal documents but is not always immediately accessible. Different repository structures and extensive policies increase search effort, especially in time-critical situations.

The RAG system makes internal documents searchable through a straightforward chat interface. Employees can describe their situation in natural language and receive a contextual answer based on approved company documents.

Every response includes references to the underlying sources, allowing employees to verify the guidance directly against the currently applicable policy.

A Next.js interface provides an accessible entry point for employees. The Python backend uses FastAPI to process documents, perform semantic retrieval, and prepare relevant context for Ollama.

The modular architecture supports operation in controlled infrastructure. PostgreSQL manages documents and metadata, while Docker enables reproducible deployment of the individual components.

Employees gain faster access to binding internal information without manually searching through folder structures and long documents. Combining a concrete response with source references supports consistent and verifiable action in daily operations.