Onyx setup: connectors, knowledge bases and agents

This tutorial covers Onyx deployment through Elestio and a walkthrough of its chat and administration interface. The speaker presents Onyx as an open source, self-hosted platform for team knowledge and LLM access. The demonstrated setup uses an OpenAI API key; it does not demonstrate local inference or offline operation.

The setup connects a GPT model and enables GPT image 1.5 using the existing OpenAI key. The walkthrough also covers web search and crawler settings. Team members can join through invitations or a CSV import. The presenter says invitations switch registration to private sign-up and shows basic and admin roles, noting the lack of finer permissions at the time of recording.

The knowledge base example indexes the English Wikipedia page for giant pandas. The presenter enters the page name rather than its URL and sets recursion to zero to limit crawling to that page. After indexing, the connector feeds a document set that the chat can query through internal search. The demonstrated answer draws on both the indexed page and an outside website.

The final section explains how an AI agent can have its own prompt, starting setup, knowledge access and actions. It also points to Slack and Discord bots, MCP actions, and an API, with API keys available through user management.