wtf is zoraos?

ZoraOS describes itself as a local-first operating system for AI workflows, aimed at research, coding, and knowledge management tasks, rather than a single chatbot. Requests move through a frontend built with Next.js, into a FastAPI gateway, then through a planner that breaks a goal into tasks, a router that picks which model or provider handles each task, and out to a set of agents and tools that do the actual work. The project ships with several named agents, including Research, Developer, Writer, and Knowledge, along with tools for the filesystem, Git, web search, running Python, and reading PDFs. It stores information using a vector store, a knowledge graph, and a document store, described together as its memory layer. A scheduler is present as a scaffold for cron-style workflows but is not yet a working executor, and a separate research daemon can run bounded research cycles locally, but only when an operator starts it directly with an exact goal, since it never launches automatically with the API. The system is designed to work with many model providers rather than being tied to one, listing OpenAI, DeepSeek, Kimi, Qwen, Anthropic, Gemini, Grok, and Ollama as supported options. The README also describes an optional multi-machine setup, splitting orchestration, inference, automation, and memory storage across separate machines. The maintainer is explicit that this is a prototype: unknown tools are denied by default, control of third-party desktops is disabled, the API is local-only, and the research daemon defaults to local models with strict daily limits on tasks, tool calls, and tokens. Getting started involves copying an example environment file, adding API keys, and running either Docker Compose or a startup script, with a separate local development server for the dashboard UI.