Nanobot setup: WebUI, local models and deployment

This tutorial covers Nanobot, a self-hosted AI agent framework written in Python and described as MIT licensed. The speaker walks through installation, browser and terminal chat, configuration, and deployment. Model options include cloud providers and local models through Ollama or vLLM; hosting the agent yourself does not establish that every provider choice keeps inference on your machine.

The stated requirement is Python 3.11 or newer. Installation options include a macOS or Linux installer, uv, and pip with the nanobot-ai package. The source checkout route also includes building the WebUI with bun or npm. The quick start opens the browser interface on localhost port 8765, then covers provider selection, model settings, and a first chat. Terminal sessions and a background gateway provide other ways to use the runtime.

Later sections explain the message bus, provider fallback chains, workspace configuration, and MCP connections. The speaker describes Dream as a background job that extracts facts from conversation history into workspace memory. Inline subagents receive separate context, tools, and provider settings, while cron tasks return results to their originating chat. Deployment coverage includes Render, Docker Compose, Linux systemd, and macOS launch agents. These are tutorial claims and setup paths, rather than independent tests of security or performance.