Vidai

Vidai functions as a dedicated control plane for managing language model requests directly within your own network perimeter. Key capabilities include: * Cost attribution for all requests * Inline policy enforcement and audit * Data sovereignty and privacy * Provider and framework agnostic operation * High performance with low latency This platform extends the traditional role of a gateway to include comprehensive governance, ensuring sensitive prompt content never leaves your network for decision-making. Written in Rust, it maintains a small footprint (25 MB in the hot path) while handling high request volumes efficiently, demonstrated by 21,803 requests per second at a median of 1.95 ms with governance actively applied. Vidai helps organizations meet stringent regulatory and compliance requirements, mapping directly to standards like ISO/IEC 42001 SoA, EU Act Art. 12, DORA, and GDPR Art. 44. Its self-hosted nature provides complete control without "phone-home" functionality, reinforcing data ownership and security. A free Community edition is available, even capable of running on low-spec hardware like a Raspberry Pi. This tool is ideal for finance teams needing to control spend, compliance officers needing to prove adherence, and engineering teams requiring safe, scalable, and latency-critical operations for their large language model deployments. It provides the essential infrastructure for secure, compliant, and cost-effective use of conversational technologies in enterprise environments.