sekacorn/linux-of-ai

Linux of AI is a collection of seven separate open-source tools designed to work together so that organizations building AI systems are not locked into one vendor or provider. The name is a nod to how Linux gave people a free, open alternative to proprietary operating systems. This project aims to do something similar for AI infrastructure, covering things like agent coordination, private deployment, model switching, policy rules, organizational meaning, audit records, and cost tracking. Each of the seven projects handles one layer of the problem. OpenOntologyLite defines business concepts and permissions outside of application code. AgentPolicyPack provides rules that govern what AI agents are allowed to do. AgentForge coordinates multiple agents, handling things like routing between models, budgets, and memory. PrivateAIStack gives a starting point for running AI locally, useful for organizations that cannot send sensitive data to outside providers. ModelSwapBench helps test whether switching to a different AI model would still deliver good results at a lower cost. AIAuditLog creates tamper resistant records of what an AI system did. AIMeter OSS measures the cost, usage, and actual business outcomes of AI systems rather than just tracking raw token counts. The README is explicit that this project is not a foundation model, not a hosted AI service, not a compliance certification, and not a legal guarantee that any model change is safe. It is meant purely as open infrastructure that people can inspect and use themselves. Each of the seven projects is already published as its own package on PyPI, and the ecosystem includes a demo showing how a team might use these tools together to evaluate replacing an expensive AI model with a cheaper one, using only local files and reports rather than needing an internet connection. The full README is longer than what was shown.