Antirion

Antirion is a high-performance, self-hosted gateway designed to manage interactions between your applications and various model providers. It's built as a single Rust binary with an integrated console. Key features include: * Provider fallback across multiple model services * Configurable budgets per team and API key * Redaction of Personally Identifiable Information (PII) and secrets * Response caching for improved efficiency * Comprehensive logging of every request, its routing, and associated cost This robust system seamlessly integrates with existing OpenAI or Anthropic SDKs, acting as a transparent layer that enhances control and observability without introducing latency. Benchmarks show it adds only 0.5 ms overhead at 5,000 requests per second (RPS) on 4 vCPUs, capable of sustaining 14,000 RPS. Antirion is ideal for development teams and enterprises seeking to optimize their use of external model services. It provides critical capabilities for cost management, data privacy, and reliability, ensuring that interactions with external models are efficient, secure, and fully auditable. Its open-source nature allows for complete transparency and customization.