OpenRAG setup: document chat, citations and MCP

OpenRAG is an open source, self-hosted document chat platform that bundles Langflow, OpenSearch and Docling. This tutorial covers deployment through Elestio, provider setup and questions across uploaded files. The presenter describes the software as free without feature gates; hosting and model usage are separate considerations.

The setup lists Anthropic, OpenAI, IBM and Ollama as LLM providers. The presenter chooses OpenAI for both generation and embeddings, enters an API key and keeps the default models. Ollama appears as an option, but the tutorial does not demonstrate a local LLM configuration or offline operation.

Using AI-generated sample documents, the presenter asks for a regional revenue breakdown, then combines an employee policy question with a product shipping specification. OpenRAG returns answers with source references. The knowledge view exposes document chunks, processing details and the model used. File selection accepts one document at a time in the demonstrated interface, while folder upload is also available.

The overview covers Drive connectors, MCP and API access, plus ingestion settings in Langflow. A practical limitation is user access: the presenter says the instance lacks invitations and team accounts. Sharing credentials or issuing API keys for MCP access are the alternatives described.