Show HN: Open-source agentic runtime – sandboxed, auditable, self-hosted (github.com)

Democr.ai has announced the beta release of its open-source agentic runtime framework, designed for building AI applications that prioritize reproducibility, auditability, and operational control. This framework encompasses a server-driven UI and integrates observability, OS-level sandboxing, and pluggable model orchestration, making it significant for the AI/ML community, particularly in environments with compliance and regulatory constraints. Unlike traditional AI stacks that require multiple components for UI, AI processing, and observability, Democr.ai consolidates these elements into a single cohesive system, aiming for production-grade deployment from the start. Key technical features include support for multi-client rendering, native multi-tenancy, a triple-layer audit system, and a strict modular architecture defined by a public SDK. The platform allows for seamless integration of various AI engines, both local and remote, while ensuring security through foundational primitives like OS-level sandboxing and RBAC. The architecture emphasizes observability and reproducibility as core design principles, providing detailed logging and audit trails for all actions and AI calls. Overall, Democr.ai positions itself as a robust alternative that avoids vendor lock-in, enabling developers to maintain full control over their AI applications.