Researchers need easier and more automated ways to create FAIR-compliant metadata. Existing tools lack coverage and consistency; ConnOSS aims to address these gaps using machine learning while supporting good research practices. The FAIR principles emphasize the importance of machine-actionable metadata for improving research quality, transparency, and reproducibility. However, researchers developing software often lack the time or expertise to manually produce comprehensive metadata and therefore seek automated, low-effort solutions. Current tools and schemas (e.g., CodeMeta, Bioschemas, maSMP) only partially cover metadata elements and face challenges with inconsistency and limited automation. There is also a clear need to harmonize metadata from multiple sources (e.g., GitHub API, citation files, README files) and make it easily accessible for both humans and machines.

The goal is to develop the Connected Open Source Software (ConnOSS) infrastructure that:

• Provides a GitHub/GitLab-based platform showcasing research software with consistent, harmonized, and enriched machine-actionable metadata.
• Improves the visibility, FAIRness, and reproducibility of research software.
• Supports researchers in FAIRifying their software with minimal effort through automation and machine learning.

• Specific schema for ConnOSS aligned with schema.org (extended via CodeMeta, Bioschemas, and maSMP).
• Machine Learning (ML) for metadata enrichment from unstructured sources such as README files.
• Metadata extraction pipelines from GitHub, GitLab, and other structured repositories.
• Web infrastructure for publishing metadata through GitHub/GitLab pages.
• FAIR and open-access practices integrated into both the infrastructure and ML models.