Nikolay Nikitin, PhD, Research Lead at the AI Institute of ITMO University, has developed OSA, an open-source Python library that uses LLM agents to automatically improve open-source repositories. The tool addresses a common problem where developers, especially in scientific research, lack time to create proper documentation, README files, and CI/CD configurations for their projects.

OSA works through a command-line interface or web GUI, supporting multiple LLM providers including OpenAI, OpenRouter, and local models via Ollama or FastAPI. The tool offers three modes: basic for standard improvements, automatic for LLM-driven analysis and recommendations, and advanced for full manual control. It can generate README files in both standard and article-style formats, create documentation, set up CI/CD workflows, and even summarize project strengths and weaknesses.

The tool was originally designed for scientific researchers in fields like biology and chemistry who often lack software engineering expertise, but it works on any repository. OSA employs an experimental multi-agent system that decomposes repository improvement into specialized tasks, handling everything from license file creation to community documentation setup. The tool supports GitHub and GitLab platforms and can adjust configuration files for proper deployment workflows.