Hey HN! I’m excited to share Aqueduct, an open-source AI gateway you can self-host to manage LLM traffic without hitting paywalls for basic features. I'm part of a small team at TU Vienna, and we wanted to provide AI services with an OpenAI-compatible API for our university researchers with zero hassle across institutes. Every AI gateway we looked at (LiteLLM, Portkey, Kong, etc.) locked critical features like SSO, organization management, observability, or self-hosting behind enterprise plans. We built Aqueduct to give our researchers a no-compromise solution without any "taxes." No SSO tax, no observability tax, no org management tax, no self-hosting tax. It uses LiteLLM Router SDK for routing (models/providers/fallbacks) on top of Django for a maintainable and hackable code base. Core features: - Usage tracking, retries, rate limiting - OpenAI-compatible API - Coming soon: MCP Server management, OpenAI Responses API, Batch API GitHub: https://github.com/tu-wien-datalab/aqueduct Thanks for taking a look! |