opennswm-lab/faros

FAROS, short for Foundation AutoResearch Operating System, is a Python-based framework that automates the process of conducting and writing up AI research. Its current focus is on research about large language models (LLMs), but the architecture is designed to be extensible to other research domains. The core idea is that instead of building one fixed "AI scientist" application, FAROS defines research as a workflow runtime. You describe a research workflow using a "blueprint" (a configuration file), specify who or what should execute each step using a "profile" and "providers" (such as an LLM API), and the system carries out the full research pipeline automatically. The currently included workflow, called ml_paper, has four stages: idea refinement (generating and ranking research ideas), experiment (scaffolding the code workspace for the experiment), paper drafting (generating a LaTeX-formatted academic paper), and reviewer simulation (producing structured feedback on the draft as if from a peer reviewer). Each stage produces artifacts, files that feed into the next stage. It is an early release candidate. Features like parallel execution, a web-based console UI, and database-backed metadata are listed as not yet included. Running it requires Python 3.11 or newer, Node.js 18 or newer, LaTeX tools for PDF generation, and a configured AI provider. The full README is longer than what was provided.